ABOUT THE INNER CIRCLE GUIDES

The Inner Circle Guides are a series of analyst reports investigating key business issues and the customer contact solutions that can help, along with various use cases, the reality of implementing and using these technologies and a view on what the future holds.

The Inner Circle Guides are free of charge to readers. Sponsors have not had influence over editorial content or analyst opinion, and readers can be assured of objectivity throughout. Any vendor views are clearly marked as such within the report.

There are Inner Circle Guides to:

  • Agent Engagement & Empowerment
  • AI-Enabled Agent Assistance
  • Chatbots, Voicebots & Conversational AI
  • Cloud based - Contact Centre Solutions
  • Customer Engagement & Personalisation
  • Customer Interaction Analytics
  • First-Contact Resolution
  • Fraud Reduction & PCI Compliance
  • Next-Generation Technology
  • Omnichannel
  • Omnichannel Workforce Optimisation
  • Outbound & Call Blending
  • Remote & Hybrid Working Contact Centre Solutions
  • Self-Service
  • Voice of the Customer.

These can be downloaded free of charge from https://www.contactbabel.com/research.

As well as explaining these solutions to the readers, we have also asked the potential users of these solutions whether they have any questions or comments to put directly to the report’s sponsor, and we have selected some of the most popular to ask. These branded Q&A elements are distributed throughout the report and give interesting insight into real-life issues.

The statistics within this report refer to the UK industry, unless stated otherwise. There is a version of this report available for download from https://www.contactbabel.com/research with equivalent US statistics and findings.

“Small” contact centres are defined in the report as having 50 or fewer agent positions; “Medium" 51-200 agent positions; and “Large" 200+ agent positions.

DRIVERS FOR SELF-SERVICE

Like so many other technology solutions in the customer contact arena, self-service started off as supporting a cost reduction strategy, and while this benefit is still very much part of the self-service tool box, there is now a growing emphasis placed upon gathering information about these interactions to feed into voice of the customer programmes, improving the customer experience and optimising the knowledge base.

As with most successful business implementations, self-service is seen as win-win for both the organisation and the customer.

The three main drivers for self-service considered in this section are:

  • Increased profitability: through avoiding unnecessary live contactsand increasing cross-selling and upselling
  • Improved performance: through easing pressure on agents, performance metrics such as speed to answer and call abandonment rates are improved, along with customer experience
  • Meeting customer demand: many modern customers wish to do business outside typical office hours, and are confident about using self-service. Having 24/7 availability on this channel means that customers can solve many issues how and when they wish, without the business having the expense of fully staffing a contact centre all day. Customers also value having a choice of channels depending on what they are trying to do: sometimes speed is of the essence, but at other times only a human interaction will do.

INCREASED PROFITABILITY

Unlike some efficiency-enhancing technologies, self-service can be beneficial to both the business and customer. Through 24/7 availability, revenues can be increased, and the cost differential between a live interaction and one carried out with self-service is enormous.

LIVE CONTACT AVOIDANCE

One of the key business drivers most quoted for self-service is contact avoidance: the reduction of the number of live interactions into the contact centre, and therefore a drop in costs. However, avoidance of contacts without fully delivering the required quality of service means that customerswill call the contact centre (or go elsewhere), so businesses must remain wary of over-extending their self-service offering.

One of the main rationales for most contact centre business investments is cost reduction, assuming that any change does not have a negative impact on the quality of service. This has certainly been the case for self-service - whether through IVR or website -where after the initial investment has been made, cost per interaction is extremely low.

Figure 1: Cost per inbound interaction (phone, social media, email & web chat) figure1

Although there is some cost differential between email, phone, social media and web chat, it is by no means as dramatic as for self-service, as there is still a relatively low level of automation being used in many businesses. For emails, it is also the case that if the query is not answered satisfactorily within a single response, the time and cost associated with multiple replies and possibly phone calls is soon greater than if the customer had simply called in the first instance.

AUTOMATED CROSS-SELLING / UPSELLING

By AI-enabling self-service, profitability may be boosted by analysing and assessing the outcomes of live and automated interactions where specific cross-selling and upselling approaches were made to customers.

Analysis can show the most successful outcomes after considering the initial reason for the interaction, the type of product under consideration, the customer type, the order of presented offers and many other variables in order to fine-tune this approach in the future.

IMPROVED PERFORMANCE

The pandemic caused the contact centre industry to undergo a crisis, with call durations, abandonment rates and speed to answer reaching levels which are nowhere near returning to anything close to the historical norm.

Figure 2: Historical average speed to answer & call abandonment rate, 2004 - 2023 figure2

Through taking away unnecessary calls, self-service can benefit the customer by relieving their frustration with the contact centre queue, and free up agents for those customers who actually need them. This is not a new benefit for self-service, but these charts show the full impact that the pandemic and its aftermath has had on performance.

Call durations have also risen to new heights, but this is less to do with the pandemic: there is no sharp upward swing in 2020 or 2021, but rather a continuation of a long-term upward trend.

This is driven in large part by the customer uptake of self-service, which means that the average call handled by an agent is more complex than ever before, leading to longer call durations.

While this impacts on businesses’ costs, it does show that self-service is making a difference in shifting low-value calls onto a much cheaper channel.

Figure 3: Historical call duration (service and sales), 2004-23 figure3

MEETING CUSTOMER DEMAND

Successful channel uptake is generally an iterative process. Businesses introduce a channel (usually based upon it being cheaper to support than the incumbent channels), and customers trial it. If it works for all concerned, it can be deemed successful. However, if customers trial it and reject it, either vehemently or simply by reverting to the existing, well-known methods of contact, then businesses have to consider whether to drop the channel quietly or amend it so that it meets the needs of the customer, who will then retry it and make their decision accordingly.

Some of the key features that customers look for in a channel include:

  • The perceived effectiveness of the channel: customers contact a business because they want something done. Feeling satisfied that their request has been taken care of is a vital ingredient to this - many contact centres still get calls asking if they have acted upon a customer’s email - and the reassurance provided by a real-time channel that an issue has been dealt with appropriately should not be underestimated. Most businesses now look to leverage their email or messaging functionality to send out a cheap outbound confirmation once a successful self-service transaction has taken place.
  • Channel availability: one of the advantages of telephony is its ubiquity. Of course, the rise of cheap computing and the popularity of smart devices now means that the telephone is under serious challenge in the availability stakes. Self-service is by its very nature a 24/7/365 channel, which is a major advantage. However, if escalation to a live agent is required, this may not always be possible outside working hours.
  • Ease of use: familiarity also comes into this. Although it may seem as though most people are comfortable using a phone or computer, some poorly-designed IVR or speech recognition systems can make life difficult for customers. Even after many years of use, customers can still be unsure about just how much a particular speech recognition system understands, may be irritated by some systems’ propensity to repeat back everything that is said to it, or are unsure if they have to wait until the speech recognition system has finished ‘speaking’ before being able to respond themselves.
  • Low cost of use: a particular issue for some demographics, with excessive amounts of time spent on hold in a queue costing a significant amount of money. In such cases a freephone number leading to a IVR self-service system may be welcomed, and of course there is no incremental cost to the customer in using web self-service either.
  • Painlessness: a customer’s subjective view on the amount of effort the overall customer interaction experience has needed, including the requirement for any follow-up interactions. There have been numerous studies done over many years about customers gripes towards IVR systems, in particular poorly-designed IVR menu structures, popularly known as ‘IVR Hell’. There is a great deal of focus being placed upon the customer effort required to deal with businesses, which is another excellent indication that the movement towards true customer-centricity is continuing within the industry.
  • Speed of conclusion: this refers to the immediacy of response and the overall resolution time, including the need for any follow-up work, or the wait-time to get an answer. Perhaps the greatest benefit to the customer of self-service, regardless of the channel or device from which it is accessed, is the fact that there is never any queue to wait in before being attended to. As surveys have consistently shown, first-contact resolution is key to customer satisfaction and this is as applicable to the self-service channel as it is to any other.

Taking these customer requirements into account, it is clear to see that self-service can offer customers what they want in many cases, especially for simpler transactional communications. Bearing in mind that self-service is far cheaper for the business as well, it is evident that self-service can be a mutually beneficial solution.

The major ongoing issue is to improve the customer experience, and the increased use of personalisation, the ability to escalate a query quickly and with context (moving between channels to live agents if necessary), and increased customer familiarity with self-service will continue to drive this channel forwards.

However, there are many suitable tasks that are not yet being dealt with through selfservice, and - perhaps more importantly for the customer experience - many which are not being handled correctly by self-service functionality.

Many of the major flaws occur when customers are forced into using self-service, or when they can’t complete their task effectively, leading to calls from unhappy customers. These negative customer experiences can be exacerbated if the history and context of what the customer was trying to do is lost in the escalation process.

Prepare for Tomorrow: Contact Centre Technology Decisions to Make Today

Business and tech trends are dynamic. Some accelerate while others lose value as more innovative solutions emerge. To make the best contact centre technology purchasing decisions, it’s important to keep up with where markets are going. Then apply that knowledge to use cases that serve your business goals.

To make informed buying decisions, make certain your long-term needs are clear and logical — and they follow customer experience trends. By tracking emerging trends in contact centre technology and how they can impact your business, you can focus on specific categories where you’ll build your best use cases.

Let’s look at some of the “big” contact centre trends in this age of artificial intelligence (AI) and the cloud.

Digital Goes Beyond Customer Self-Service

Many businesses start their digital journeys with self-service capabilities. Today, consumers use an average of almost six touchpoints with nearly 50% regularly using more than four. Your digital customer experience strategy needs to be connected and cohesive to give customers a consistent experience — and that consistency includes voice.

These connections enable you to transform experiences with personalisation and then optimise those experiences based on contextual understanding of who’s engaging with you — and why.

Look for Built-In Use Cases That Simplify AI

At its core, the value of AI is in the wide range of capabilities it enables that support innovation. This includes conversational AI services with generative AI for automation that also enables a human touch. Using this embedded power of data and automation, businesses can create new products and services, streamline operations, and enhance the experiences of customers and employees.

As you explore AI capabilities, predictive AI can equip your businesses with deeper customer insights for personalisation and accurate forecasts to improve workforce planning. And because implementation and maintenance are simplified with embedded capabilities, AI reduces overall costs and complexity.

Customer Data Must Be Protected and Available

Establishing trust is vital. You and your customers should feel confident that the AI systems you use will protect their data and privacy, while operating reliably, accurately and ethically. Designing for transparency and trust aligns with customer perceptions of your brand. When they see the benefits of sharing their data in the form of more personalised experiences, it builds trust and loyalty.

Ask the Right Questions About the Right Customer Experience Trends

Digging into the specific capabilities — and promises — that vendors offer helps ensure that the customer experience you deliver today uses the best contact centre technology. That innovation should continue as the foundation of differentiating your services for many years to come.

For more relevant questions to ask vendors about how their technology will support your long-term plans, read our 2024 contact centre buyer’s guide.

Read the full blog here - Prepare for Tomorrow: Contact Centre Technology Decisions to Make Today

CHANNEL PREFERENCES

To gauge the level of customer demand for self-service in various scenarios, a survey of 1,000 UK consumers was carried out to understand what the channels of preference would be in cases of high emotion, urgency and complexity through presenting survey respondents with three hypothetical scenarios:

High emotion: e.g. notifying a company that an incorrect item has been sent to them. This was chosen as a high emotion interaction as being sent an incorrect item is often frustrating, as not only has the desired product not arrived, but the customer is then left with the problem and effort of returning the item. This is not a particularly complex interaction, and in many cases will not be particularly urgent.

High urgency: e.g. checking the arrival time of a flight that the customer is meeting. This is likely to be an urgent interaction as it is very time-sensitive. Complexity is very low -as the required information is simply a time - and in the majority of cases, should have a fairly low emotional impact.

High complexity: e.g. receiving guidance on completing a mortgage application or tax form. This is likely to be a complex and long interaction, but is unlikely to have high levels of urgency or emotional response.

HIGH EMOTION INTERACTIONS

Consumers taking the survey were asked to imagine that a product they had ordered from a company had arrived but was incorrect. In this circumstance, they were asked which would be their preferred method for contacting the company to notify them that this was the case.

The most popular option was to email the organisation, with 32% of respondents choosing this method. The second most popular, at 24%, was phoning the contact centre, and web chat also made a strong appearance, with 16% respondents choosing this as their preference.

There was a strong pattern based on the age of the survey respondent and their preferred channel: the older demographics were the most likely to pick up the phone or to email.

Web chat was a very popular option with the younger demographics, and has actually overtaken the telephony channel for the 16-24 year-old cohort.

13% of the sub-25 year-old age group would choose social media, which is a major finding for businesses serving these customers.

Figure 4: Preferred method for contacting a company (high emotion interaction), by age range figure4

As this research has been running annually for six years, it now has enough history to look at how channel preferences are changing.

The pandemic has clearly made a significant (and seemingly lasting change) to how customers prefer to interact with businesses.

This is perhaps least dramatically shown in the chart below, which shows how customers prefer to handle high emotion interactions. Although there has been a drop in the proportion of customers choosing the email channel, there has not been the attendant increase in telephony that can be seen in other interaction types, although there has been a rise in the preference for face-to-face interactions.

It should be noted that research for the customer survey is usually carried out in early Q2 of each year, and that 2020’s figures consequently do not show the full impact of the first pandemic lockdown. Statistics from 2021 onward are taken from customers who have experienced the decline in service levels that was undeniably present in many companies due to the pandemic.

Figure 5: Change in channel preference (high emotion interaction), 2018-23 figure5

HIGH URGENCY INTERACTIONS

Survey respondents were asked which would be their preferred channel of choice in a situation where they were meeting somebody from a plane and needed to confirm the time at which to be at the airport.

The most popular channel was that of telephony, with most age groups choosing this as their no.1 option. This is quite a change on pre-pandemic findings, which put web selfservice as clearly the most popular channel of choice, and may be a reflection of customers’ greater requirement for the reassurance and confidence that the phone channel provides.

Email, social media and web chat were more likely to be preferred by younger demographics, but not exclusively by any means.

Figure 6: Preferred method for contacting a company (high urgency interaction), by age range figure6

The effects of the pandemic can be seen clearly: live telephony has replaced web selfservice as the preferred channel for urgent interactions, despite the massive investments put in place by many businesses to achieve the opposite effect.

It is not possible to state with complete confidence why this should be, but it may be that many customers have experienced very poor levels of customer experience from some companies that struggled in the pandemic and afterwards, and that they have reverted to the channel that they associate with confidence, flexibility and resolution: telephony.

Figure 7: Change in channel preference (high urgency interaction), 2018-23 figure7

HIGH COMPLEXITY INTERACTIONS

For highly complex interactions, such as getting expert guidance with a tax form or mortgage application, the most popular contact choice pre-pandemic had been making a physical visit to an office or branch, which was much more popular with the older demographic.

However, this option has dropped in popularity, probably due to the customers getting out of the habit of making unnecessary visits during the pandemic, particularly as the experience would likely to be different than what they are used to.

It might have been expected that the next most-personal channel would have grown in popularity as a result, and telephony has risen from 16% in 2018 to 36% in 2023.

Web chat was also seen as an appropriate primary channel for complex interactions by a significant minority of 25-54 year-olds, whereas email is generally much less popular than it is for high emotion interactions, possibly due to the probable requirement for back-and-forth communication.

Figure 8: Preferred method for contacting a company (high complexity interaction), by age range figure8

As with urgent requests, the preference for telephony jumped hugely during the pandemic and has remained high, probably for the same reasons.

Web self-service has seen the greatest drop in preference, with customers preferring to be reassured by an actual person: this is borne out by the relatively low drop in face-to-face communication, as if this was simply a matter of not wanting to risk these situations for health purposes, then telephony would have replaced face-to-face.

It seems as though customers - possibly through their own unsatisfactory experiences -have formed an opinion that they simply want to be talked through their complex issue rather than try and fail to do so using self-service.

These figures suggest that self-service still has some way to go before customers choose this channel by preference.

Figure 9: Change in channel preference (high complexity interaction), 2018-23 figure9

CHANNEL USAGE

Looking at how channel usage is changing, as not all of the same respondents take part in this survey every year, a jump or drop in the usage of a minor multimedia channel could be an industry-wide phenomenon or a case of a handful of early adopters skewing the results, which is certainly possible where only a few use a channel, and where mean averages are used.

As such, a question is asked to respondents about how each inbound channel will change, so being able to judge if any alterations in the use of channels is due to real changes at a contact centre-level, or is more of a statistical blip caused by a different set of respondents providing data each year.

Figure 10: How do you think inbound channels will change in your contact centre in the next 12 months? figure10

As usual, the traditional media of letters will have a net decline in our respondents' eyes, although this still has a small place in the likes of the insurance, medical and manufacturing industries. More respondents believed the live telephony channel volumes would drop (54%) than thought they would rise (20%), a finding that has grown for some years. This trend in the industry is explored in the next chart.

Strong growth is once again expected in web chat interactions, with SMS / messaging and social media also predicted growth. However, after many years of expected growth, respondents expect email volumes to stay fair flat. Telephony self-service is expected to grow once again this year, with its twin benefits of customer convenience and low cost still very much relevant. New approaches, such as visual IVR, could encourage further use of self-service. Although not shown on this chart, around half of respondents offer an app or mobile service option for customer service.

The previous chart's real message is that channels aren't being replaced, but rather augmented, and businesses have to accept that they need to develop an omnichannel approach, as that’s what their customers are expecting. This means that the pressure to unify the view of the customer across channels is a challenge that isn't going to go away.

The following chart shows a historical representation of answers to this question, showing how the enthusiasm and expectation of channels has changed. Respondents could choose one of five options connected with how they believed each channel would grow in the next 12 months, and a score was given to each to reflect its effect: greatly increase (+2); slightly increase (+1); no change (0); slightly decrease (-1); or greatly decrease (-2). This would give a net score of between -200 and 200, with positive scores expecting growth and negative scores decline. For example, a channel where 70% expected a slight increase and 30% a slight decrease would receive a score of +40 (i.e. “70” + “-30”).

Figure 11: Net expectations for channel change in next 12 months, 2011-23 figure11

Web chat shows very strong ongoing growth, having net scores of 100 or more for most years since 2012. However, social media outlook is less positive this year. Email has shown a distinct cooling in expectations since 2012, and dropped into negative territory (i.e. an expected decline) in 2020. It is expected to show no net growth in the next 12 months.

The expectations for live agent telephony had dropped considerably since 2015, showing a definite decline in its relative importance although its rate of expected decline is lower again this year. Telephony self-service is generally expected to grow, and letters to continue their decline.

Looking at the reality of omnichannel activity, the UK contact centre industry has now strongly embraced the various forms of non-voice customer communication.

Figure 12: Inbound interactions by channel figure12

The proportion of live inbound interactions by telephone is 64.1%, in line with a long-term gentle downward trend which appears to have stabilised in recent years, despite businesses’ expectations that phone volumes will decline.

The proportion of telephony self-service interactions rose from 4.9% in 2019 to 6.2% in 2020 and 2021, in line with the growth of self-service across other channels, but dropped again in 2022 and 2023.

The email channel increased significantly in 2017 to around 20%, after being around 15% for a number of years. It is still the largest digital channel at 19.4%.

Web chat has grown strongly, but appears to have fallen back a little this year. However, further strong growth is expected by businesses and the impact of AI-Enabled chatbots has only just been begun to be felt. Social media’s figure has returned to being below 2% after a brief period around the 5% mark.

SMS / messaging steadies at 1.8%, with growth being driven in large part by a rise in WhatsApp / Messenger rather than standard text messages. This is up from 1.2% in 2021.

Looking at vertical market figures, agent-handled calls are most important to respondents in the housing and insurance sectors, with manufacturing respondents (usually working in B2B environments) and retail (which uses email and web chat) reporting lower levels of telephony as usual.

Figure 13: Inbound interactions by channel, by vertical market figure13

Email is well represented in most vertical markets, with the manufacturing, retail and services sectors the highest.

Telephony self-service seems strongest in the utilities sector as usual, and the public sector respondents report a higher-than-usual finding this year.

Web chat is developing a very strong presence in retail, so as to encourage and close online sales, as well as handling service queries.

The manufacturing sector reports being ahead in terms of social media customer contact, although this may be a one-off statistical blip.

24/7/365 IMMEDIATE SERVICE

Having looked at when customers are most likely to use self-service, this section considers what customers believe makes for a good customer experience and how the judicious use of self-service can support this.

ContactBabel commissioned the research firm Aurora Market Research to carry out a survey of 1,000 UK consumers. One of the purposes was to identify any differences in opinion between organisations and customers about what were the most important customer experience factors when contacting an organisation.

As such, consumers were asked to state which were the top three most important factors to them when contacting an organisation.

Figures below are expressed as the percentage of each age group that expressed an opinion.

Figure 14: What are the top 3 most important factors to you when contacting an organisation by phone or digital channel? (by age range) figure14

The previous chart shows the importance of various customer experience factors as an aggregated bar chart, segmented by age so as to show the factors that were of most importance to customers in each age range. Aggregating the results allows an understanding of which factors were placed in the top three overall, while also providing insight on age-related opinion.

For example, 41% of the youngest age group (16 to 24 years old) stated that having a short wait time for a response was one of their top three most important factors, whereas 56% of the oldest age group (over 65 years old) placed this in their top three.

This consumer research has some interesting findings when comparing consumer attitudes to businesses’ beliefs:

  • both businesses and consumers agree that first-contact resolution and short wait times are the most important factors impacting upon customer experience when contacting a business
  • long opening hours are seen as more important by customers than businesses
  • having UK-based employees is seen as far more important to customers than businesses believe, particularly for the older generation.

When considering these findings from the perspective of the various age ranges, the importance of first-contact resolution is much more popular across all age ranges, which is a change from previous years when older demographics felt this to be much more important.

For most of these key customer experience factors, self-service - when used effectively -can give the customers want they most value. It is worth noting that younger customers place more importance than the older generation on longer opening hours, and are also more likely to value having a choice of ways to communicate with the organisation.

Further evidence for this age group’s valuing of its time can be seen in relatively high importance being placed upon short call/web chat duration compared to the older generation and suggests that the younger generations are potentially keenest to use selfservice.

CURRENT AND FUTURE USE OF SELF-SERVICE

Self-service is found across most industries-there is often at least one function that selfservice is suitable for, regardless of what a company actually does - but some sectors use it more than others.

Figure 15: Some functions for self-service, by vertical market figure15

END-USER QUESTION #1 : WHICH SELF-SERVICE SOLUTIONS ARE APPROPRIATE AND AFFORDABLE FOR SMALLER CONTACT CENTRES OR THOSE WITH RESTRICTED BUDGETS?

Self-service experiences are a great way to scale customer service. The technology allows an organisation to do more with less, reaching more customers, across more time zones and across more channels. Self-service is a great way for companies with smaller contact centres to achieve scale without requiring a large agent headcount.

The ideal solution for smaller contact centres is one that embeds AI, minimising the need to find, fund and integrate additional capabilities. This embedded AI should include capabilities such as NLP (natural language processing) to help the solution understand customer inquiries, it should include predictive and prescriptive routing capabilities to support personalisation, and AI-Enabled knowledge to serve as a backbone of information. Al should be easy to use. Administrators should have AI-Enabled tools to enable Al-driven customer self-service - smart search, bots, etc. and also AI-Enabled agent tools. Agents in a small contact centre are especially challenged to do more with less and often have to manually wade through massive amounts of information. AI can help streamline this in a variety of ways. This process will speed up the average response time for agents while also allowing them to draw upon a central source of truth, in this case a knowledge base.

89% of respondents use offer some form of self-service to customers, with FAQs / general search being available to 75%, and account-specific web self-service to 61%. The former allows a search of the site as a whole, perhaps using FAQs or text search, whereas the account-specific variety requires a customer login in order to access functionality and information specific to that customer.

Touchtone (DTMF) IVR is used widely across size bands, and as with those using automated speech recognition, those in the largest operations are more likely to use it. A similar finding applies to virtual agents/ chatbots. A small proportion of respondents in large operations use visual IVR. Crowdsourcing is used by 22% of respondents in large contact centres.

Figure 16: Use of self-service, by contact centre size figure16

WEB SELF-SERVICE

For businesses, by far the major advantage to having customers use web self-service is the fact that the cost per automated support session is estimated to be between 40 and 100 times cheaper than a live call to an agent.

Research has found that around 50-60% of calls to the contact centre result from bad website service or a failure in another channel. Quite apart from the current importance of this application, research shows that as customers become more educated and experience many different qualities of online self-service, their expectations increase across the board which puts pressure on other organisations to keep up or even exceed the current benchmark performance.

Most customers will visit a website first; if they cannot find what they’re looking for immediately they will try self-service; if the self-service experience does not give them what they want immediately and accurately, they will either call the business or go elsewhere. In cases where the customer is tied into an existing business, this will result (merely) in a higher cost of service and decreased customer satisfaction.

In cases where the web visitor is only a potential customer, a failure in the self-service process on a website will mean the a I most-certain loss of a sale. In all cases, providing effective web self-service options-with a clear path to escalation to a live agent, along with any contextual customer specific information - is in the best interests of the business.

In terms of pure self-service, the website can provide various options for the customer, ranging from the most basic search and static FAQ functionality, to personalised virtual agents and dynamic FAQs.

By far the most prevalent form of web self-service is that of the FAQ (frequently-asked question), which is used by 75% of respondents. The free text search of the document library is somewhat less well supported, at 20%. Virtual agents are employed by 37% of respondents, more often those within larger enterprises.

SEARCH

Since corporate websites first came into being, businesses have offered search tools for customers to look through indexed information, based on keywords found in these documents, in order to answer their questions without the need to call the business.

While such functionality has the advantage of at least being familiar, indices grow, documents get old and out-of-date, and customers become educated that there are more sophisticated and effective self-service solutions available, with customers’ opinions of standard search functionality suffering as a result.

With only a blank text entry box to guide them, the onus to search successfully is with the customer, who has to try to ‘get into the mind of the business’ and phrase the question or search terms in a way that fits the business and its internal jargon.

However, this is not always possible, and customers have a limit to the maximum number of times that they will attempt to search, or how many pages they will read from the numerous documents that a wide keyword search can bring back, claiming that it has answered the query. The customer then has two possibilities: to engage the business through a high cost channel such as telephony or email, or worse, to find an alternative supplier that can help them without going through this high effort process.

Search functionality does have its place: for example, if a customer wanted to find out very specific information about a product that had an unambiguous name (for example, ‘SDK36479 installation’), a search on this particular term would at least bring back documents that had a high level of relevance to this product and how to set it up.

However, if the customer had a query that used keywords that were very popular and widely found elsewhere (for example, “What are your delivery times?”), typical search functionality might return every document that contains the word ‘delivery’, relying upon the customer’s patience and goodwill to find the correct answer for themselves. In the case of very large companies, this could bring back potentially hundreds or thousands of documents, many of which could be out-of-date and have been superseded. The major problem with search functionality is that it pays close attention to the answers, but very little to understanding the question or the customer’s thought processes.

It is one thing to be presented with a long list of documents while sitting in front of a large screen of a PC, where scrolling up and down the page is not an issue. For the same flawed search functionality to be placed onto a mobile website, expecting the user to zoom in and out, scrolling up and down, and then to potentially scan through numerous documents whose text is too small to read properly is probably a step too far even for the most enthusiastic and loyal of your customers.

Some self-service solutions alleviate this issue by using customer feedback to judge the success of the search results provided so as to increase future customers’ chances of being given the correct information.

FAQS

FAQs-frequently asked questions - are one of the most popular forms of web selfservice. At its simplest, an FAQ list can simply be a group of static documents and/or text, categorised under wider thematic headings, and kept up-to-date manually. Solution providers state that perhaps 80% of questions can be answered by 20% of documents, however for most businesses, customer requirements change on an ongoing basis so it is unlikely to be the same 20% of documents that are most useful as time progresses.

More complex applications can use techniques such as text mining and fuzzy search (approximate string matching) to return documents that are not just an exact or very close match to the search terms entered by the user. Sophisticated FAQ technology will leverage natural language processing to deliver more accuracy than standard search functionality.

It is possible to minimise the use of manual updates and supervision by making the FAQ list more dynamic and self-learning through using responses taken from emails to customers who have asked specific questions, which will then dynamically enter the FAQ list at an appropriately high level. Being able to restructure the knowledge base on a regular and ongoing basis through automation is key to maintaining the usefulness and relevance of the FAQs.

Unlike the virtual agent (below), FAQs by their nature provide the user with a list of alternatives, asking them to judge and choose the correct most relevant answer for themselves. While this process takes longer for the customer than the provision of a single answer, it is currently more closely aligned with the typical user experience, and thus has the advantage of familiarity. Providers of FAQ technology report that the typical reduction seen by customers in inbound live contact (such as email or telephony) is in the region of 25%.

VIRTUAL AGENTS AND CHATBOTS

AI in the customer contact world is perhaps currently best known for chatbots, also known as virtual agents. Conversational AI tools such as natural language processing (NLP), machine learning and analytics run automated tasks and simulate conversation with the customers.

The chatbot program using Conversational AI may be given a human avatar and personality characteristics, and can include natural language processing, dialogue control, access to knowledge bases and a visual appearance that can change depending on who it is talking to and the subject of the conversation. Customers communicate with these applications using natural language rather than simply keyword recognition, and the bots reply using context and in a manner that a human would.

Chatbots - as the name suggests - are found in the web chat channel, but the functionality can be used in any other digital channel, such as social media, email and especially voice self-service in the form of voicebots. Through being trained on large amounts of data and using machine learning and natural language processing (NLP), AI-Enabled chatbots can recognise user speech and text along with context and intent. A key part of Conversational AI is the ability to improve itself over time through a constant feedback loop which improve the AI’s algorithms.

However, chatbots do not always use Conversational AI, and rule-based basic chat applications can be used in circumstances where there are a significant proportion of interactions about similar issues, where the chatbot can be tasked to answer these and pass more complex issues to a human agent. They are trained to answer only a preprogrammed, specific set of questions, and while they can be extremely useful, are not seen as part of the Conversational AI world. They cannot handle queries outside of their programmed scope, and do not use machine learning to improve their responses. Any changes to their output or flow needs to be done manually.

Even without AI, rule-based chatbots can have a positive impact on customer care without requiring large investments or long implementation timescales. Chatbots using Conversational AI use natural language understanding (NLU) and can ask questions to understand customer intent and improve the accuracy of the output, and may also use machine learning and analytics to predict and improve future outcomes.

Web chat-whether through chatbots of whatever description, or live agents - has experienced significant growth and although volumes on average are still less than 10% of all customer/business interactions, in some vertical markets they are considerably higher. There is no reason why the user uptake of web chat will not continue: it works well for customers as it provides a quick response, and with multiple concurrent chat sessions per agent, it can be a lower cost channel than voice for the business to support. This cost differential is getting particularly noticeable as there has been a significant movement towards the use of chatbots.

While the cost of web chat is dropping, there is still considerable room for increasing efficiencies and lowering costs.

Whereas only 5% of web chats had any automation involved in 2015, this had grown to 53% in 2023, mainly as a result of initial handling by automated chatbots which may then hand off to live agents where appropriate, although fully-automated AI-Enabled web chat has increased very significantly in recent years as well.

Figure 17: Level of automation used in web chat, 2015-23 figure17

Virtual agent functionality ’understands’ the context of what the customer is asking, with the result being more akin to that of an empathetic human who also has had access to what the customer has been trying to do. For example, if asked “When can I expect my delivery?", the context and the required answer will be different depending on whether the customer has placed an order and is enquiring about its status, or has only a hypothetical interest in turnaround times in case they decide to place an order.

When the virtual agent application has low confidence that it has returned the correct result, it is able to escalate the customers query seamlessly to a live chat agent, who then has access to the self-service session history, enabling a greater chance of a successful resolution without repetition. (It is generally considered best practice that escalations to real agents are not hidden from customers). The eventual correct response can be fed back to the automated virtual agent and the knowledge base underlying it, which will make it more likely that future similar requests can be handled successfully through automated agents.

MOBILITY & SMARTPHONES

Statistics that show the number of smartphone users, volume of apps downloaded and the value of mobile transactions are rising so quickly that they would be out-of-date before this report is published. It is sufficient to note that with very few exceptions, the mobile customer is relevant to every organisation, in every vertical market, in every geography of the world.

The rapidly decreasing cost of mobile bandwidth, coupled with the huge improvements in mobile networks mean that businesses can be ambitious in what they are attempting within this channel, having an opportunity to build presence and functionality in an area that is growing rapidly.

On average, 73% of calls received by a contact centre are made on a mobile phone.

84% of survey respondents state that more than half of the calls made to their operation are done through mobile phones rather than landlines, offering huge potential for valueadd services such as video, visual IVR and other mobile-related functionality.

Figure 18: What % of your customers use a mobile phone to call a contact centre? 2016-23 figure18

At a vertical market level, respondents in utilities, public sector, transport & travel and housing report the highest proportion of calls being made from mobile phones.

Research shows that 91% of customers who have a poor experience with shopping on a mobile site will abandon it: some may intend to return via a PC, but many others will search elsewhere: there is no differentiation or allowances made for sub-optimal mobile web experiences.

Furthermore, most businesses are currently failing in this attempt, with the mobile channel lagging way behind online websites and bricks-and-mortar shops. Offering a mobile customer experience tends to mean offering a smartphone app and/or a mobile version of a website.

Mobile Websites

A mobile website differs from simply accessing a full website via a mobile browser, rather offering a mobile-optimised alternative which is easier to use and overcomes some of the constraints around using a smartphone to access the web, such as tiny fonts, excessive scrolling and difficult-to-press buttons.

Mobile websites usually do not try to offer every single item available on the full website, but focus upon the information and processes that most users will want in order to act or make a decision. Ease of use is vital: text must be fully displayed on screen, buttons must be clickable and businesses have had to consider minimising the use of graphics to achieve quicker load times in areas with poor mobile data services, although this is becoming less of an issue as fast and cheap mobile data becomes more widespread.

Bearing in mind that a mobile site generally cannot support every type of interaction that a customer may want, businesses may consider that allowing mobile users to access the main website is a good idea. Contact details should be clear, and offering a seamless route from self-service into supported service, via email, web chat or telephony is very desirable.

It is beneficial for businesses to understand why customers are using a mobile site rather than waiting until they are in front of a PC: the request may be related to what they are doing at that current time, and so waiting is not appropriate. Generally, customers will be more task-focused on a mobile device than a PC, so the emphasis should be on delivering quick, simple, high-volume interactions, i.e. those ideally suited to self-service.

For example, by looking at the current use of their full website, a bank may discover that a high proportion of users want to check their bank balance or view recent transactions rather than setting up automatic bill payments or ordering foreign currency.

Consequently, the mobile version of the website may focus only on a small number of the most popular high-volume interaction types.

Smartphone apps

A good app may provide a superior user experience to a mobile website, due to the greater level of design. However, they tend to be much more expensive to build, and unlike a mobile website, a new one has to be developed for each smartphone platform. Additionally, company apps will tend to be free to download, so there is little opportunity to make money directly from them. However, for many businesses, the cost savings made by having customers self-serve via an app rather than calling the contact centre are very considerable.

Smartphone platform market shares show that Android and iOS shipments account for almost all of the market, so businesses could decide to produce only two flavours of app, which would actually support the vast majority of the smartphone market.

A native application developed for a mobile device can use some of the device’s capabilities to enhance the customer experience. For example, a smartphone app can prompt drivers at the scene of a car accident to provide and capture the correct information, including photos. Such an app could also use GPS to give the exact location of the accident for use by the insurance company.

Industry estimates for building an app vary considerably depending on what they are trying to do, but many sources indicate that a cost of £20,000 upwards (per platform) is very feasible. The cost of developing a mobile website is less, and only needs to be done once. Whether an app is suitable for a company depends on their budget, and their customer base. It may be that the superior branding associated with apps is seen as being well worth the expense, even before factors like increased sales conversion rates and reduced live contact costs are taken into account.

62% of this year’s survey respondents stated that they offer mobile functionality for customer service, with a further 19% having plans to do so.

Respondents from large operations are somewhat more likely than small and medium contact centres to be using mobile customer contact functionality already.

Figure 19: Use of mobile functionality (app, mobile website) for customer service, by contact centre size figure19

VISUAL IVR AND VIDEO

VISUAL IVR

The audio-only nature of DTMF IVR places limitations upon how user-friendly the experience can be for a customer. There has always been a trade-off required between functionality and usability, which manifests itself in the number of menu options and levels that made available within the IVR system.

The rapid growth in smartphones has meant that it is now possible to offer a visual representation of IVR menus on a device which will then be used to call the business. Because it is far quicker to read text than to listen to text being spoken - some studies show that a caller can navigate a visual IVR menu between four and five times quicker than a DTMF IVR menu - the customer experience is improved without sacrificing any functionality or options. Furthermore, visual IVR can be used to send video presentations while waiting for an agent, for educational or marketing purposes, or to answer the selfservice requirement (for example, pushing the relevant YouTube clip in order to show the caller how to do something).

Many businesses that use DTMF IVR have made long-term investments in this technology, and retiring the system entirely is not desirable. Giving existing IVR functionality a visual interface simply means that the IVR’s path can be shown as a picture on a website or smartphone, with callers touching the selection that they require without having to listen to all of the options or to go up and down levels or branches. This has the dual benefit for the customer of being far quicker than listening to IVR menu options, and of being significantly more likely to get them the correct information or to be routed to the department most appropriate to their needs. Visual IVR menu systems integrate with existing DTMF structures and reuse the same VoiceXML scripts, meaning that any changes made to the existing DTMF IVR system will be automatically replicated regardless of channel or device.

Visual IVR offers companies the ability to develop value-added applications for their customers, rather than simply providing a visual representation of existing IVR menus. For example, in cases where very specific expertise is required, visual IVR can be used to help the caller self-diagnose where in the organisation they need to be going, rather than having to speak to a front-line agent who will then have to ask them the same questions in order to route the call to the appropriate resource.

It is worth noting that despite the huge uptake in smartphones and mobile apps, it is very unlikely that customers will find it convenient to have an app for every company with which they deal. Like apps, a visual IVR option provides businesses with an opportunity to display corporate branding and deliver an improved customer interaction experience.

Figure 20: Visual IVR: benefits for businesses and customers figure20

Building a business case for visual IVR may involve looking at the self-service ‘zero-out’ rate for your specific industry compared to your own statistics, considering your call transfer rate and listening to the ‘Voice of the Customer’ via call recording or speech analytics as they comment upon their IVR experience.

Carrying out a specific IVR customer experience survey is also a good way of gaining accurate insight into what might turn out to be a significantly negative experience for some of your customer base.

VIDEO

Visual IVR (some similar functionality is also referred to as IWR - interactive voice and video response) provides both visual and audio information, allowing businesses to send video clips for educational or marketing purposes to mobile telephones while waiting for an agent, or to answer the self-service requirement (for example, presenting the relevant YouTube clip in order to show the caller how to do something). Self-service videos on the website are an effective way to support technical issues as it is far easier for someone to understand a solution when they are shown it, rather than being told about it.

While video is often thought of in the context of showing a company’s agents in a live environment, it can be used as part of a supported multimedia service experience, with the agent sending relevant recorded video clips either via chat or email, or simply having them available through self-service on the website. This can improve the customer experience as well as reducing avoidable contact.

Having live or recorded video of a product on a website can be much more effective for sales and service than simply having static images, in that it can provide a full 360 degree view and zoom in on anything of particular interest. This is a technique that many car sales businesses have adopted in lockdown which is sure to continue in the future as it reduces the customer effort considerably and provides the opportunity to expand their potential market far beyond the local area.

Research also seems to suggest that recorded video works very well for fashion brands where the look of the item (rather than its actual utility) is the main driver for purchase. Recorded videos on the website have been shown to increase consumer confidence, lengthen the interaction with the website, and increase actual sales.

Analytics is used on many companies’ websites to deliver a personalised experience to a prospect or customer who’s browsing the site, through offering support based on the outcome of previous site visitors’ behaviour e.g. popping up a web chat session if they pause, hover over a site element, visit a page repeatedly, etc. It is also possible to add a pop-up that provides self-service rather than a contact option: for example, a visitor who is spending a long time trying to change an existing booking can be sent a video showing them how to do this

TELEPHONY SELF-SERVICE

Despite the rapid growth in the use of web-based services, the importance of the voice channel has not diminished to the extent predicted by some commentators, as customers still find voice the most convenient, flexible and quickest communication channel in many instances, especially in older demographics and for complex and high-emotion enquiries.

The challenge for businesses is to improve the customer experience, protect their customers’ private and personal information and control their own costs. As such, the use of automated voice-based solutions has become widespread and offers a rapid service option to customers while keeping contact centre costs down.

Voice self-service is usually delivered either by touchtone (known as DTMF: dual tone, multi-frequency) IVR, which allows customers with a touchtone phone to access and provide information in a numerical format. Some businesses, often with large contact centres and high call volumes, use automated speech recognition (ASR), which allows customers to speak their requirements to the system, allowing greater flexibility and functionality. The emergence of visual IVR - a front-end developed for smartphones which bridges the gap between digital and voice - has the potential to give self-service a significant boost although current usage is low.

IVR (interactive voice response) - whether through DTMF or speech recognition - has four main functions:

  1. to route calls to the right person or department (e.g. “Press 1 for sales, or 2 for service...’’) in auto-attendant mode
  2. to identify who’s calling via either caller-line identity (where the caller’s number is recognised, and their records brought up immediately), or through inputted information, such as account number. The caller’s information is then “popped" onto the screen of an agent who then understands who the customer is and what they are likely to want
  3. to segment and differentiate between customers, prioritising against business rules in order to deliver a premium standard of service to them (e.g. minimising time on-hold, spending longer on the phone with them, offering high-value services, etc.)
  4. to deliver a total customer service interaction without having to use a human agent, saving the business money - historically, it has been calculated that 6 or 7 self-service IVR calls cost about the same as a single person-to-person call.

This section of the report considers the role of IVR and speech recognition as part of a full telephony self-service solution, i.e. one that takes the place of an agent to handle the whole interaction.

Figure 21: Advantages and disadvantages of telephony self-service figure21

Of those contact centres offering telephony self-service, a mean average of 22% were handled entirely by self-service without requiring an agent.

Figure 22: Overall proportion of calls handled entirely through self-service (only in respondents which offer telephony self-service) figure22

Many calls are not suitable for self-service, as they may require multiple requests within the same call, be of a complex nature or be from a caller who feels that they need to speak with a person. Additionally, some small businesses may have such a low volume of calls that it is not cost-effective to implement self-service.

Even amongst those respondents for whom telephony self-service is a vital part of the customer contact strategy, it’s no use trying to shift every customer service interaction onto telephony self-service, as if customers don’t want to use IVR, they will “zero-out" (press 0 for a live agent, or try to find a similar shortcut). And if businesses don’t offer a live agent option to an irate and frustrated caller, they won’t need to worry about providing customer service to them in the future, as they’ll go elsewhere.

It is worth reiterating that if callers agree to try a company’s self-service system rather than insisting upon talking to an agent, there is an implied contract that if the self-service session is unsuitable, the caller should be allowed to speak with an agent. Few things can frustrate callers more than being hectored into using an unhelpful and irrelevant selfservice system.

DTMF (TOUCHTONE) IVR

Pure software IVR platforms run on standard servers, reducing the restrictions that proprietary hardware placed upon functionality, scalability and flexibility, as well as the cost of purchasing and maintaining dedicated hardware. Companies increasingly prefer to adopt the cloud-based method of providing IVR options to the customers, and 73% of those using telephony self-service (whether touchtone DTMF IVR or automated speech recognition) access this functionality in the cloud, with a further 23% planning to do so by the end of 2025.

Speech-enabling IVR increases the features available to the caller. Standards-based languages such as CCXML and VoiceXML support speech recognition and improved access to relevant corporate data, the integration of which into the IVR experience supports text-to-speech and the use of caller profiling to enable personalised IVR sessions based on who the caller is, their history, their contact preferences and any other relevant information that would further assist the self-service session.

DTMF IVR has the advantage of extreme simplicity for customers, which means that it may well have an important role to play on a sector-specific basis, even with the advent of newer and more sophisticated solutions. In situations where callers need the same piece of information on a recurring basis - such as checking the balance of prepaid credit cards - customers can access the information within a few seconds by typing in the DTMF digit sequence that they have learnt off-by-heart, and it may well be that this method of accessing information is the most convenient and quickest for customers.

In addition, interactions that require a simple list of digits may be more suited to the unambiguous nature of DTMF (which, unlike speech recognition, is unaffected by background noise). Of course, by far the most common application for delivering long sequences of numbers is through making a payment via credit card, and placing a customer call into an automated DTMF session in order to do this has numerous advantages for businesses and customers in terms of convenience, familiarity and security.

The take-up of cloud-based IVR solutions, particularly by small/medium sized companies, is driving growth within this sector. The ability to personalise IVR sessions, as well as the low initial start-up costs and limited in-house maintenance required, means that businesses that traditionally were unable or unwilling to see the benefits of IVR for their own company are now revisiting this.

Many solution providers state that they are actively increasing the power and range of the analytics solutions not just within live contact channels such as chat and voice, but also within automated IVR environments as well. This can be used to adapt and personalise the IVR experience in real-time to suit the customer’s behaviour and preferences, and also to detect and manage fraud.

The use of DTMF IVR (touchtone IVR) is particularly prevalent in high-volume, transactional environments. Despite relatively little planned new growth in this area, there is still significant scope to upgrade or replace this functionality, with Visual IVR potentially providing a major boost amongst smartphone users.

Figure 23: Current technology usage and short-term investment plans - DTMF IVR by vertical market, end-2023 figure23

The greater the complexity of a DTMF IVR system, the more difficult and long-winded it is for the customer to use, and businesses have to find the right balance.

Looking at the number of levels used on a DTMF IVR (i.e. how many key-presses a caller must make to reach their destination), only 28% of respondents keep it simple with a single-level of options, e.g. "Press 1 for Sales; 2 for Service; 3 for Accounts".

24% of large operations present a possible four or more routing menu levels to their customers, a level of granularity that must appear daunting to their customer base.

Figure 24: DTMF IVR routing menu levels, by contact centre size figure24

It is not just the number of levels in a menu that can frustrate customers, but also the overall number of options within each level. As the customer cannot see what the options are, but has to listen to each, it can be a very frustrating experience, and one which the movement to visual channels such as web self-service or visual IVR via a smartphone will go a long way towards alleviating.

Respondents report a median of between 6 and 7 options, which can still be a considerable number for a caller to listen to, especially if their preferred choice is the last one in line.

Logically, larger contact centres will tend to support larger businesses, which usually have more departments, offer a greater level of segmentation and have more products and services available to customers. Consequently, there are on average many more menu choices offered in the phone menu of large contact centres, with 70% of these respondents reporting offering seven or more routing options to their customers.

Figure 25: DTMF IVR total routing options, by contact centre size figure25

AUTOMATED SPEECH RECOGNITION

DTMF IVR has been a notable success for many businesses, and many have leveraged the added flexibility and power of speech recognition as well as being able to share the functionality that businesses have recently developed with their web self-service applications. Of course, this is likely to come at an additional cost, and trying to find capital budget to invest in these solutions may be difficult, leading to a significant take-up of cloud-based speech recognition.

One of the most consistently strong inhibitors against the uptake of speech recognition is the initial cost involved, as well as the expected ongoing support costs, and cloud has a particular appeal to organisations who don't wish to invest or tie-up large sums of upfront capital investment on their own systems or software, or pay for the in-house IT resource to run them.

One advantage of cloud is that the need for significant upfront technology investment is lessened, providing on-tap access to extensive telephony resource, albeit of a third-party nature. Additionally, the use of cloud-based solutions means that businesses don't need continual ongoing investment to upgrade their own systems.

Like other self-service applications, automated speech has of course been more attractive for organisations with high volumes, where the cost of handling the call can even exceed the business value it represents. In this scenario, the need to reduce cost is imperative, but for speech-based self-service to work well, the technology infrastructure on which it depends must be robust enough, and the number of phone lines linked to it large enough to accommodate the maximum number of callers ever likely to contact the service, or run the risk of turning callers away, an opportunity cost which can be very high.

Cloud-based speech services, where the telephony and technology infrastructure is centrally owned and managed by a third party overcomes this capital investment hurdle, and the pay-as-you-go model adopted by most cloud suppliers means that ongoing operating costs are directly pegged to transaction volume, providing valuable operational flexibility.

Despite the wider and more powerful functionality that speech recognition gives to an IVR system, significant inhibitors are present. It is generally acknowledged that speech recognition can be considerably more expensive to implement than DTMF IVR, and is also likely to require significant in-house resource to fine-tune and operate it going forward.

Some solution providers note that the majority of businesses’ interest in moving from DTMF to speech recognition comes when the existing telephony self-service legacy system is approaching end-of-life.

Speech-based IVR is particularly useful in cases where very long lists of items such as place names or surnames may be chosen, for which the more structured DTMF IVR is unsuited. The success or otherwise of speech-based IVRs is very affected by how callers are encouraged to use the service.

Some speech implementations have actually made life more difficult for the customer, who may lack confidence that the system will understand them and so provide very short, one-word answers; if nothing is given in the way of prompts or examples, callers may give too little or too much information as they are unsure of the sophistication or capabilities of the system, and this can lead to high self-service abandonment rates. Using prompts such as “describe in a few words why you are calling us, for example 'to start a new mortgage application’" can be extremely useful in setting ground rules for the successful use of the system.

As there are so many speech recognition applications now in use in daily life -for example Siri, Alexa and PC-based voice recognition software - consumers are now becoming more comfortable giving voice commands to an automated system. With every successful speech interaction, customers’ confidence increases and speech-enabled self-service becomes a little more firmly embedded in the customer base’s psyche.

As with DTMF IVR, the use of automated speech recognition is a factor of call volumes and the level of transactional contact. As it also tends to be significantly more expensive, it is far more prevalent in large operations. As penetration into large operations is already significant, the greatest CAGR will occur in mid-sized operations.

Figure 26: Current technology usage and short-term investment plans - speech recognition by vertical market, end-2023 figure26

VOICEBOTS

A voicebot is an application made up from AI and natural language understanding (NLU). Voicebots convert speech to text, analyse it and respond appropriately using text-to-speech. It is integrated with CRM or a knowledge base in order to provide a greater accuracy and depth of response. It should be noted that a common use of speech recognition, such as keyword spotting in order to route a call, is not the same as a voicebot.

Like any machine learning application, voicebots require training, as well as volumes of clean data from which to learn. Voicebots are used to deliver full self-service experiences without requiring an agent, but are also increasingly used for customer identity verification and also call routing.

An AI-Enabled voicebot can have conversations with customers to determine even multiple intents and deliver a far wider range of personalised information and services. Crucially, it can learn from experience.

Advantages and use cases for voicebots include:

  • Cost reduction - a typical self-service voicebot interaction costs around 30-50p per call, compared to over £5-6 for a live agent
  • By asking questions and understanding the content and the context of what the customer requires, call routing can be far more quick and effective than touchtone IVR or keyword-driven automated speech recognition. In some cases, it is possible to turn a routing request into a self-service session, avoiding the need for a live agent at all
  • Providing 24/7 service in multiple languages
  • Create far wider opportunities for extending self-service capabilities through greater sophistication and understanding of the customer’s intent
  • Taking pressure off agents, particularly in times ofcrisis or volume surges
  • PCI DSS compliance can be easier, voicebots can handle card payments rather than needing agents to do this, without customers having to type in numbers on a keypad
  • Inbound and outbound activity can be linked to provide superior customer service: for example, an outbound SMS appointment reminder may initiate a change request from the customer which can be handled by a voicebot
  • Encouraging customers to use natural language generates large amounts of data that can be used to further train AI models
  • Voicebots can be used for outbound work as well, such as debt collection, reminders and surveys
  • Personalising the voicebot allows greeting by name, and a change of voice/ speaking style depending on customer preference. Customer identity verification is also possible.

While screen popping is useful for cutting time from the early part of a call, the insight that this functionality provides is often limited to the customer’s name and a general idea about why they are calling.

AI-Enabled voicebots offer instantaneous gathering and assessment of data from multiple sources to occur even before the call has been routed, which allows accurate prioritisation and delivery of the call if a live agent is needed, or initiate a self-service session if appropriate.

For example, a voicebot working in an airline contact centre may judge a call to be urgent if the caller:

  • Has booked a flight for this day
  • Rarely calls the contact centre, preferring to use self-service
  • Is a frequent flier
  • Is calling from a mobile phone rather than a landline
  • Shares a similar profile with other customers who only tend to call for very urgent reasons.

In such a case, the AI may consider that there is a likelihood that the call is directly related to the flight that is happening today (e.g. there’s a danger of missing the flight and the customer may need to rebook), and after listening to the customer’s (possibly anxious and stressed) request, can either help through self-service or move the call to the front of the queue, routing it to an agent experienced in changing flights and whose communication style suits the situation and customer profile.

ТНЕ BUILDING BLOCKS OF SELF-SERVICE

This section of the report looks at the practicalities of implementing self-service solutions and is divided into four parts:

  • Making sure self-service is suitable for your business
  • The role of AI IN SELF-SERVICE
  • How to implement self-service and avoid pitfalls.

SUITABILITY OF SELF-SERVICE

Customers choose to speak to a business through the particular channel which they believe best suits their requirements, which are usually quite generic regardless of the actual query. Specifically, customers look for service experiences which are:

  • Effective
  • Quick
  • Low effort
  • Cost-effective.

If a channel does not meet each of these requirements to a significant extent, it is unlikely to succeed.

The majority of customer interactions fall into one of two categories: those that are purely transactional and those that require dialogue (interactional):

  • Transactional communications, such as balance enquiries and travel information, require access to highly structured business information, and non-automated transactions can require an agent to act simply as an 'organic interface' between the back office systems and the customer. Such communications are suitable for self-service, whether through DTMF IVR, speech recognition or web-based selfservice options which offer the speed and flexibility of visual information.

    By putting an automated front-end on top of an existing back-office process, cost per transaction is reduced very significantly. Customers value the speed of the selfservice transaction compared to the alternatives (telephony, face-to-face, letter, etc.), so this works well for both businesses and customers. However, when customers have questions that require help in order to complete a transaction, they need to be able to get the answers, either by escalation to an agent (along with the context and history of what they’ve been trying to do), and/or through access to a company’s knowledge base.
  • Interactional communications, such as technical helpdesks, complex or multiple enquiries or where the customer requires reassurance and confirmation often require actual dialogue and discussion between the customer and the business's representative.

    It is important to note that it is not solely the level of complexity that drives a customer to choose live contact over automation, but also the state of mind of the customer at that time. For example, a customer may value reassurance rather than speed in certain circumstances (e.g. wanting to check train times to go to a wedding, or making an important hospital appointment). In such cases, not allowing the customer to interact with a live agent will have a considerable negative impact on their opinion of the organisation, potentially far outweighing the extra cost that is associated with a single instance of providing a person to talk with, rather than a self-service option. Additionally, some customers simply prefer speaking to another person and even the best self-service application is anathema to them. Results of customer surveys around channel preference by type of interaction were shown earlier in this report.

Suitability of self-service can be seen as a function of the complexity and volume of interactions.

Figure 27: The effect of complexity and volume on the use of self-service figure27

The previous diagram suggests that the greater the number of simple interactions a company deals with, the more likely it is that it can benefit from implementing selfservice.

  • Large volumes of simple requests from customers (and who use agents simply as a means of reading the information from a screen - little more than an 'organic interface’) should have implemented self-service by now. Historically, there were estimates that 70% of calls to helpdesks were password reset requests, which selfservice was able to handle.
  • Where businesses only deal in a relatively small number of complex interactions, the cost of implementing a sophisticated self-service application - and keeping the knowledge base up-to-date - may be greater than any associated salary cost reduction.
  • Businesses having a small number of simple interactions may choose to have simple voice self-service functionality hosted in the cloud, paying perhaps only for the number of times that it is used. This model allows self-service functionality at a fraction of the cost of owning and maintaining a premises-based system. Alternatively, website FAQs can deflect a significant proportion of calls.
  • Businesses which deal with large numbers of complex interactions are building and using some of the most interesting and potentially beneficial self-service applications, particularly with the assistance of artificial intelligence.

    Examples include filling in insurance quotes - a lengthy and time-consuming business, costing the business a great deal of money. Moving this to self-service can save huge amounts of money, as an agent may only need to be brought in to close the sale or clarify finer points of the policy and customer familiarity with price comparison sites has made this a much more widely used type of self-service. The next level for self-service is to use AI and machine learning to converse in a natural way with customers, asking questions and clarifying responses to provide a service close to being comparable to a live agent.

There are two main factors that influence contact centres within any vertical market: the commercial activity within that sector, and customers’ requirements and preferences for contacting organisations.

It is not only the nature of the specific business vertical market that needs to be considered. The urgency, complexity and emotional importance of the interaction is perhaps at least as important as the nature of the business that is being called: for a customer calling a bank, a simple balance request and an urgent call about the progress of a mortgage application are very different types of call, and should be treated as such.

How customers view each case can be seen in the earlier part of this report, “Channel Preferences”.

The Customer Interaction Cube (below) is a structure developed to categorise the different types of customer interactions that businesses have to handle, considering the urgency, complexity and emotional input of the interaction from the customer’s perspective.

Businesses could use this to analyse their volumes of each type of interaction, crossreferencing it with other variables such as the time of day these types of interaction are received, and the customer demographic preferences seen next in this section in order to support the relevant channels through the promotion of alternatives to live calls, and the correct levels of resourcing.

Doing this will not only improve the customer experience, but also reduce the cost of service through anticipating the likely resourcing required and even proactively engaging with the customer on lower cost channels first.

cube

Using this 2x2x2 cube as a structure, there are eight types of interaction, a combination of either low or high urgency, complexity and emotional input. Our hypothesis is that each of these eight interaction types may best be suited to specific channels, and that both business and customer could benefit from matching channel with interaction type.

The examples shown below of various scenarios and the channels most suitable for these are suggestions, and will differ between customer types, businesses and vertical markets, but may offer a tentative framework for readers to build their own scenarios. It should be noted that the results of the customer survey that follow this section suggest that different age groups and socioeconomic segments have their own views on how they prefer to contact a business in each of these cases. Primary and secondary channels are suggested here, but will differ between organisations and customer types.

Figure 28: The Customer Interaction Cube and suggested associated channels figure28

There are many other variables that could be considered alongside these that will impact upon the suitability of channels:

  • Demographics
  • Ownership of smartphone / broadband impacts upon channel availability
  • Time of day (i.e. is this an out-of-hours enquiry? Is the customer at home, at work, or travelling?)
  • Whether the request is specific to an account, or a generic issue (i.e. is it necessary to passthrough security first?).

While the 2x2x2 cube can help businesses to estimate the current and potential volumes and resourcing required to serve the customer base, it is important to remember that similar types of customer interaction may require very different handling depending on circumstances.

For example, a query about product delivery may be a small part of a wide-ranging research process carried out by a particularly thorough prospective customer, or may be asked by a customer who has just realised they’ve forgotten about an important birthday and needs immediate, accurate information.

THE ROLE OF AI IN SELF-SERVICE

Artificial intelligence (AI) is a wide-ranging term for technology solutions which appears to emulate human cognitive capabilities through the 'understanding' of complex, natural language requirements, in order to reach its own conclusions and develop itself based on what works and what doesn’t. Machine learning refers to the ability of software to evolve based on measuring its performance and success, without input from humans.

Within the customer contact space, there is a great deal of interest in how AI can work to deliver a superior customer experience at every hour of the day, across channels, leveraging the vast amounts of data that are available to many large organisations. Supported by the speed and availability of affordable processing power, and the enormous amount of structured and unstructured data available, the opportunity exists for AI to take customer contact far beyond what is feasible now.

Although we are towards the beginning of the AI revolution, there are already numerous well-known examples widely used by the public, including Amazon’s Alexa and Apple’s Siri. These virtual assistants 'understand' unstructured natural language requests and deliver the solutions in a manner similar to a live personal assistant.

As AI can be given access to all of the relevant data a company holds on its customers, as well as unstructured data held elsewhere (for example, forums or social media channels), it has a far wider source of knowledge from which to draw, compared to human agents. In theory, an Al with sufficient sophistication could make human agents all but unnecessary, but for the foreseeable future, AI will usually work alongside its human colleagues.

AI DEFINITIONS

AI (artificial intelligence)

Within the contact centre, AI involves technologies such as machine learning, speech-to-text, deep learning, analytics, chatbots and natural language understanding, all closely integrated and working together, aiming to provide outcomes similar or even superior to those achievable by human agents.

Some of the typical characteristics of AI-Enabled solutions include:

  • An understanding of the customer’s meaning and intent, rather than just accurately decoding the syntax of the request
  • Use of multiple questions in a conversational format to improve understanding
  • Using past outcomes to predict and deliver the likeliest most successful output
  • The use of confidence levels rather than a binary right/wrong output
  • The ability to improve future outcomes without constant human input or monitoring.

Chatbots / Virtual Agents / Virtual Assistants / Conversational AI

As with so much in the world of AI, there is disagreement about definitions. In the case of chatbots, virtual agents, virtual assistants and Conversational AI, it is better to focus on the functionality and ‘intelligence’ powering it, rather than the phrase in itself.

AI for customer contact is currently best known for chatbots, applications that run automated tasks and simulate conversation with the customers. It may be given a human avatar and personality characteristics, and includes natural language processing, dialogue control, access to knowledge bases and a visual appearance that can change depending on who it is talking to, and the subject of the conversation. Chatbots are often found in the web chat channel, but the functionality can be used in any other digital channel, such as social media, email or even voice self-service.

Chatbots are not always fully-automated or AI-Enabled, and may in fact be a glorified FAQ interface, lacking ‘understanding’ and simply searching through keywords. However, some use NLP and can ask questions to understand customer intent and improve the accuracy of the output, and may also use machine learning to improve future outcomes.

In this report, “chatbots" and “virtual agents” are used interchangeably and refer to the same functionality.

Virtual assistants (VAs) are not dedicated to a single task (such as customer service), and can assist in numerous ways such as taking notes, carrying out web research, setting alarms, communicating with smart devices, etc.

Both chatbots and VAs are conversational interfaces, but the level of AI involved can differ greatly.

Machine Learning / Deep Learning / Neural Networks

Through the use of pattern recognition, previous outcomes and other algorithms, machine learning enables systems to improve themselves without the need for continuous human user input (although supervision and guidance is often needed in reality). It relies upon extensive datasets and computational power in order to make predictions with theoretically continually-improving levels of confidence.

Based on the workings of the human brain, neural networks consist of input and output layers as well as one or multiple hidden layers (Deep Learning uses multiple layers, each carrying out their own specific task), working to find patterns which will be too onerous or complex the humans to identify. Neural networks can be trained to spot patterns in data and provide accurate output, with programmers correcting any mistakes. Eventually the neural network can ‘understand’ whether it is producing accurate output, with far less human correction.

Neural networks can be set up using supervised or unsupervised learning techniques. Supervised learning techniques involve giving the neural network a specific problem such as “is this customer likely to complain?”. Programmers then provide the system with large datasets of customers who have or have not complained, and then the neural network will find patterns of characteristics that make some customers more prone to complaint. They are then able to predict which customers are likely to be dissatisfied, allowing the business to act accordingly. In the case of unsupervised learning, no specific output is given to the system, which will then find patterns in the data and classify groups accordingly. Supervised learning is by far the more common use of AI in businesses.

Natural Language Processing / Understanding (NLP/NLU)

NLP refers to the branch of AI which enables computers to understand human language, whether spoken or written. It goes beyond speech to text processing - although of course accurate transcription is vital - and attempts to understand the actual intent of the customer. NLU is a subset of NLP which looks at the challenges of understanding human communication, such as mispronunciation, sub-optimal word order, slang and other elements which are a natural part of human speech but which can cause major problems for computers due to their unstructured and outlying nature.

One of the keys to successful automated service, whether via telephony or website, is for the user to be able to describe their issue in their own words, rather than feeling that they have to use specific terms or a stilted, incomplete account of the issue. Natural language processing-based systems encourage users to describe their issue more fully, asking follow-up questions if there is any degree of ambiguity in the initial request.

One of the obstacles to overcome for NLP-based systems (whether through speech recognition or text recognition) is that many Internet users have been trained to use keywords, believing that simplifying the description of their issue will lead to greater levels of accurate response. In fact, NLP works best with longerand more detailed requests, and it is a challenge for businesses and solution providers to encourage and support users of the system in using the solution in an optimal way.

OPPORTUNITIES FOR AI-ENABLED SELF-SERVICE

Improve Voice Self-Service

Using AI-Enabled natural language recognition can alleviate the high level of self-service abandonment associated with speech recognition and DTMF IVR, as there is no fixed menu to navigate and no limit to the number of options a customer has to explain their issue. The onus is placed upon the system to understand the customer’s intent, rather than forcing the customer to shoehorn their request into a format allowed by the predefined rules and format of the business.

Improve Web Self-Service

For most businesses, the customer is given free rein to search through documents, prewritten answers and archives, hoping to stumble across the right answer for themselves. This often proves time-consuming and ultimately frustrating for the customer, who will then go elsewhere or call the contact centre in a negative mindset. An AI guide would be a valuable aid in improving CX and deflecting unnecessary calls.

Assisted Service

The use of AI to assist agents in real time within a call offers the chance of a real paradigm change: by the nature of the job, an agent-customer interaction has always necessarily been between two people, and the level of support that an agent can actually receive within a call is very limited. Al can work alongside agents to provide relevant knowledge that may be otherwise take a long time to find, and update the knowledge bases available to humans and AI self-service systems using an automated feedback loop that is constantly improving based on actual outcomes.

Improve Digital Channel Experience and Decrease Cost

Perhaps the currently most popular use of AI in the customer contact environment is in handling digital enquiries: web chats generally take far longer than phone calls (due to agent multitasking, and typing time) and some email response rates can still be measured in days. The next section looks at this in some depth.

As the cost of web chat is broadly similar to other channels such as email, voice and social media, there considerable opportunity to increase efficiencies and lower costs. Digital channels may work well for customers, but businesses are not generally seeing the cost savings that automation can bring.

Very few emails or web chats are handled entirely by AI, although a growing proportion of web chats are dealt with by Ais working alongside agents, suggesting responses which agents can then accept or amend. This way of working is most likely to be the norm in the foreseeable future, with the speed of automation and the emotional intelligence of humans combining to provide superior service at a lower cost.

KNOWLEDGE BASES

One of the most central and critical elements to a company’s service capability is the knowledge base, which is vital to the accuracy and consistency of the AI-Enabled selfservice experience for both agent and customer across channels.

For many organisations, a knowledge base started off as a list of useful documents and files, which quickly grew into a wider, less coherent collection of information sources, requiring increased levels of expert management, amendments, editing, and deletion.

However, the resources required to keep these knowledge bases up-to-date are very scarce, as the people within the business that have the capabilities and expertise to do so also have their own jobs to do. Very quickly, what started off as a useful and highly tailored information resource has mushroomed into an expensive, out-of-date and increasingly less-useful collection of information of wildly varying quality. AI can assist in the management of knowledge bases by feeding back successful outcomes, and noting when the answers provided did not meet the requirement.

On an ongoing basis, feedback from agents and customers will identify gaps in the knowledge base which will need to be filled by product experts. Some knowledge bases will require full-time, dedicated resource to manage them, whereas others will rely on automated systems making dynamic changes depending on callers’ and agents’ requirements.

Often, the challenge is collecting the right knowledge from across the organisation and turning it into articles with structured, visual content. Therefore, start by focusing on the knowledge required to meet the highest-value use cases, which have a large volume of interactions, a small number of possible responses and very clear searchable tags. Bear in mind that this does require dedicated staff and a clear delivery plan.

It is often the case that large businesses with many products and services to maintain will have numerous editors across many departments who can make suggestions, although it may only be a small handful of people who will verify and publish this information.

Businesses may want to consider allowing certain contact centre agents to create new entries based on their communications with the customer. Understanding which documents are being used the most allows the maintenance efforts be focused on the most important areas.

It is not just the publishing of information that is vital: crowd-sourcing of answers, and feedback on accuracy and success from the wider “super-user" community will help the business to fine-tune the knowledge base and train any Al being used. Processes to gather this feedback should be put in place, and continually revisited to check effectiveness, and it’s possible to add successful answers to the knowledge base very quickly if a response from an agent (for example, via email or web chat) has been marked to be successful, and Al is an effective method of doing this regularly and consistently.

Those who contribute timely and useful information - whether a customer or an employee - can be rewarded and recognised accordingly. People want to share their knowledge with others, and enabling them to do so easily is beneficial for all parties concerned. Businesses could measure the success of the knowledge management system by measuring the return on investment from call avoidance, by the rating or score given by readers of recommended articles, or through customer satisfaction ratings.

The process of assembling the data and knowledge can be done through data labelling, which requires a tag to be put against each knowledge source (e.g. text, pictures, videos), showing what it is about, for example “a video clip showing how to change an oil filter on a specific car model”.

As this can require a great deal of resource, another method may be to crowdsource the collation and tagging of data form a number of sources: the agent as they go about their everyday business; a field technician solving the customer’s issue; or super-customers who are happy to answer queries from other customers on a web forum.

Businesses interested in how AI can help service should aim for a symbiotic relationship between customer self-service and agent assistance, the focal point of each being a knowledge base which is continually refreshed, amended and added to by agent, customers, super-users and Al itself.

Depending on its sophistication, the creation, uptake and maintenance of a knowledge base may require a dedicated team, at least in its initial phase, of a user experience designer, data scientist and developer to build the model, with inputs from business experts to keep the model aligned with what the commercial requirements actually are.

Those looking to implement use cases which are tightly focused upon specific high-volume queries and processes (e.g. chatbots), will need less intense support. Solution providers may have editable templates and predefined applications for many popular business processes, or even have pre-trained bots. Key to success is remembering that this is about solving a business issue, not implementing impressive technology, so it is vital that both the user interface and implementation procedure are friendly for those other than AI specialists.

Once the knowledge base has been created, success largely depends on the usefulness of the information and how often it is accessed. This is different to quality and accuracy: there could be completely accurate information, but if it is rarely used, it isn’t delivering value.

END-USER QUESTION #2 : WHAT DOES CREATING AND MAINTAINING A SELF-SERVICE KNOWLEDGE BASE ACTUALLY INVOLVE?

When creating a self-service knowledge base, it’s best to start with the customer. That means being able to see what customers are requesting most often. It is also an opportunity to check in with agents. Are agents struggling to answer some questions more so than others? This insight should inform what content needs to be created. Often, the content is in the form of an FAQ but it could be articles. This is information that organisations likely already have authored, but it may be locked in a hard to access content system. This will act as a foundation for the self-service experience. While it can be beneficial to try and load the system with many possible answers, it should also be accepted that unconsidered questions will be asked. As a result, part of the initial process should involve heavily monitoring the performance of the self-service experience with an eye on gaps and addressing these with new entries in the knowledge base.

In terms of maintaining a self-service knowledge base, there are two components. One is a proactive component in that an organisation must be diligent when they're introducing a new service, promotion or change to existing policy. In these scenarios they need to ensure that the information is going into the knowledge base. If there's anything that contradicts what’s in there, they need to also go in to make sure that they're correcting that information as well. With a solution like Genesys DX, we make it very easy for brands to search the knowledge base and very quickly update it. That new information is then automatically reflected across all the channels, across all the agent interfaces, et cetera.

Now the other component is on the reactive side. For this, an organisation needs to respond to data regarding their knowledge base to optimise it. For example, let’s say they are running a timed promotion that lets customers book travel accommodations at a discounted rate for the remainder of the month. However, maybe a customer is curious if they can use that promotion to book a trip for next year, way in advance. Is that covered by the promotion? It’s possible it’s a scenario that wasn’t even considered. Either way this could identify a gap in the knowledge base that could then be addressed after finding this unanswered query. With Genesys DX, we have a workflow where you can segment data by intervals, like the past two days, past 24 hours, et cetera. So, if a change was made to the knowledge base, a good best practice can be to monitor the system to identify and quickly address questions coming through without an appropriate entry in the knowledge base. Even if no changes were made, though, it’s a good best practice to look over the system just to identify gaps and address them. The work involved doing this differs greatly, with a larger workload expected when a knowledge base is new or a significant change has occurred, and then less when the system has been running for a while without any significant changes.

HANDLING SELF-SERVICE EXCEPTIONS

No matter how good the self-service solution is, there will always be times when customers decide that they need to speak with an agent. In some cases, the self-service will be at fault, but in many, this is a decision made by the customer based on their specific needs and emotions at the time. This section looks at how often and why this happens.

WEB SELF-SERVICE

Although around one-third of survey respondents state that fewer than 10% of their customers have tried to resolve issues online before calling the contact centre, 30% state that more than 1 in 4 of their inbound calls come from people who have failed to complete their objective on the website first, and who may approach the call in a state of frustration.

Figure 29: Proportion of callers that have tried to answer own queries through web self-service before calling figure29

One of the two most important reasons for moving from web self-service to live telephony was that the escalation involved a complex issue requiring a live agent to complete successfully.

88% of respondents also felt that customers wanted the reassurance that a live agent brings to a conversation.

64% stated that the functionality that the customer calling in required was not available online, but interestingly, 63% stated that they received calls about issues that could in theory be resolved online, but customers were unable or unwilling to do so. As such, businesses may consider that time spent educating customers in how to use self-service would pay benefits in the long term.

Relatively few respondents believed that website security authentication was an issue causing unnecessary inbound calls.

Figure 30: Why customers move from web self-service to live telephony figure30

VOICE SELF-SERVICE

Overall, a mean average of 14% of calls that go into the self-service option are “zeroed-out”: instances where the customer decides that they in fact wish to speak with an operator.

There is a broadly positive correlation between the size of the contact centre and the proportion of self-service sessions that are abandoned in favour of speaking to an agent: the larger the contact centre, the more often customers ‘zero out’.

One possible reason for this might be that larger operations are trying to do too much with their self-service. There is some evidence to suggest that this is the case, as it is very noticeable that respondents from larger organisations tend to have far more options in the auto-attendant functionality of their IVR solution, and this tendency to offer a great deal of functionality and options may well also apply to IVR’s self-service functionality as well. Overly complex or long-winded IVR functionality will tend to encourage session abandonment, and this may well be what we see here.

Figure 31: Reasons for abandoning telephony self-service sessions figure31

71% of respondents agreed that customers abandoned self-service sessions because the self-service function simply does not offer what the customers want.

While this at first glance may appear negative, it is the case that even in the most commoditised and transaction-driven environments a substantial proportion of customers will want to speak to a person, either because the system does not allow them to do what they want, there is a complicating factor involved, or simply that they wish reassurance or have multiple questions.

In such circumstances, it is the customer’s choice to abandon the session, and this does not have to be a particularly negative experience as long as a clear exit path that leads to a live agent is marked early in the process. Situations where businesses hide their agents from customers, making them go around in IVR loops are the ones that give all telephony self-service a bad name.

Only 18% of respondents strongly agree that having too many options presented to customers is a major reason for them seeking human assistance. It is noticeable that 72% of respondents agree to some extent that the customer simply does not trust the system, preferring to have human reassurance that the request they have made has been carried out, or the information they are looking for is actually correct.

Of those using automated speech recognition, 36% of respondents agree that speech recognition is unpopular with customers due to lack of accuracy and user-friendliness. While this is high, it is a major improvement on past years, and it may be that customers are gaining confidence in how to use the system after many years of struggling. As customers continue to be encouraged to use natural language (both by successful interactions with corporate self-service applications, but perhaps more importantly through digital virtual assistants such as Siri and Alexa), this issue should further decline.

CUSTOMER INHIBITIONS

UK organisations and consumers were asked whether automation or human assistance would be preferable to the customer base in circumstances where the customer effort, time and outcome were exactly the same. Although the question is quite hypothetical -automated channels usually require far less effort and time than human interaction, but often cannot deliver the same functionality-the findings were quite surprising considering the popularity and recent uptake of automated self-service.

Looking at the age group of the customer base, older demographics feel more strongly about human contact, with only 4% of over-65s preferring to use automation, compared to 21% of 16-24 year-olds. This fits in with the previous findings that this section of the customer base places more value on their time, whereas the older demographic prefers to have their issue resolved first-time by a single employee. Having said that, every age group expressed a strong preference to speak with a human agent.

Bearing in mind that this question emphasised that the outcome and customer effort/time would be identical in each case, the results show that the customer base at present is not yet at a stage where automation is generally seen as being even on equal terms with human contact, let alone the preferred method of contact with a business, and that the human touch is still very much valued.

Figure 32: Would you prefer to speak with an agent or use automation, if the outcome and time were identical? (by age range) figure32

Further analysis of this data showed that 69% of men preferred to speak with a person, compared to 73% of women.

At a socio-economic level, there was virtually no difference between respondent sets.

This reluctance to embrace automation is a major barrier for businesses to overcome. Customers’ experience of self-service applications will have an impact on all of their future engagements with self-service, even if they are with different companies.

In the same way that customers were ‘inoculated’ against using the email channel in the early 2000s due to its widespread ineffectiveness, the entire customer contact industry is responsible for the wholesale uptake or rejection of next generation AI-Enabled selfservice as a preferred primary channel.

To succeed, self-service must be either quicker (i.e. less effort) than talking to an agent, or be more effective. The latter is very unlikely, so it seems the future success or otherwise of self-service will be determined by whether zero friction, easy-to-use applications become mainstream.

IMPLEMENTING SELF-SERVICE SUCCESSFULLY

To encourage the uptake and long-term use of self-service, consider a few actions which make self-service work both for businesses and customers:

Identify the quick wins

Measure the most successful self-service areas of the website, based on number of page visits and success rates at providing answers, for example frequently-asked questions (FAQs) about products, order, and delivery status, or requesting a return. These areas are prime candidates for self-service, and often align closely with the routine interactions that contact centre agents currently handle.

Start simply

Build natural language capability into self-service via speech-enabled IVR, voicebots and/or chatbots. To manage cost and time, focus on a specific set of use cases with clear customer intent, a small number of well-defined responses, and a large number of interactions. If the natural language engine does not recognise the request, it simply redirects to traditional interaction handling. Even this simple step could have a high hit rate.

Keep consistency

Leverage the same interfaces (APIs) used by the website through your voice IVR system, using text-to-speech and pre-recorded information to provide self-service information for customers calling in.

Always make it easy to escalate to a live agent. If an agent does get these type of requests, train them to access the self-service information through the IVR while the caller is listening in. This increases customer comfort levels and encourages them to use selfservice in the future.

Knowledge is power

Make sure you have a process in place where the knowledge base is continually updated and kept fresh. Issues can change quickly, especially if there is an unexpected issue that has recently arisen: updating the knowledge base should be a priority in this case.

Keeping a close eye on what customers are asking to the self-service application is a good way of gauging what’s happening in the minds of the customers.

Gather feedback from customers about how well their questions have been answered, and use this to identify gaps (but do so sparingly, and only when you will actually use this data to improve). It is often the case that in the search for simplicity, the self-service system will be set up to provide only basic answers which only go part of the way to answering the question: consider adding more detail to this by a ‘read more’ button or links to other answers.

Make self-service the first port-of-call, not the last

Publicise the new self-service functionality through other channels: the IVR announcement; by having agents tell callers about it at the end of the conversation; by promotion on the website itself, perhaps being placed more prominently and accessible with less effort than the telephone number or email address. Strong branding and an attractive, clean user interface are also key to recognition and long-term uptake, making the website a digital front door to your business.

Experience the full customer journey

Before committing to adding or improving self-service to a process, take the time to understand fully how the customer experiences not only the activity that you are looking to optimise, but also the various routes and actions that have brought them there in the first place. It may be that the typical customer has had to try other channels first, or that the current self-service functionality is hidden away or has an unfriendly user interface. In such cases, direct replacement of the functionality will not improve the play rate or success rate of the self-service action anywhere near as much as it could be.

Train your customers

For web chat conversations with customers, train the agent to share a link to the website self-service page and then walk the customer through the process. Over time, more people will choose self-service if they are confident that they know it can help them.

Encourage the use of natural language

If you are implementing a self-service system that works best with natural language inputs (rather than a simple keyword search), support customers to use it correctly. While some businesses prefer to let customers know that they are speaking with AI, this can be off-putting for some who will react by making things as simple as they can, thinking that they are helping the system that way. A customer who thinks that they are communicating with another person will do so in complete sentences, so some businesses may prefer to hide the nature of the AI-Enabled self-service to see if the customers’ inputs are of higher quality that way.

Measure success

Each business should know which outcomes it wants to see from an AI-Enabled selfservice implementation: a drop in call volumes; an improvement in customer experience scores; more sales through the website etc. It’s important to measure the factors that affect the business’s success, rather than the factors that are internal to the self-service application itself (such as the number of knowledge articles that an agent contributes, rather than beneficial metrics such as the article’s rating by customers or the reuse of each article). As with telephony agents, measuring the wrong self-service metrics encourages behaviours that may not align with what you’re trying to do with your business.

IMPLEMENTING AI-ENABLED SELF-SERVICE

It’s important for businesses to understand that if they’re not already using AI-Enabled self-service, then they haven’t already missed the boat, and that even with unlimited budget and resource, there are many contact centre activities that are more appropriate for a person to do.

For first-time Al deployments, the focus should be on delivering a high-quality self-service solution for a relatively small and clearly defined business process or issue, rather than taking on more complex situations, even if there is a potentially higher benefit. It might be appropriate to start with a tightly focused AI-Enabled self-service project, and then look to roll this out for more complex customer self-service requirements, as well as to other parts of the customer journey including call routing, back-office processing, analytics and agent assistance.

Apart from the dangers associated with an overly complex initial project, scale is also an issue to consider. To begin with, businesses may consider it wise to limit the number of concurrent customer or agent users that AI supports (i.e. dozens rather than hundreds of concurrent users), in order to learn what works and what needs improvement in each use case, and in order to optimise processing performance by providing the right amount of processing capacity.

Over time, machine learning tends to require less processor power and running a relatively small scale AI implementation for a few months will provide a more informed view of what full-scale usage of AI will involve, meaning that the right amount (and cost) of processing power can be established.

If you’re considering implementing AI-Enabled self-service, there are some questions that you should ask yourself first:

  • Is there a specific pain point or issue within the operation that needs to be addressed? e.g. lack of available resource to handle existing enquiries, suboptimal business processes, inability to analyse large datasets, etc.
  • How does this affect the customer experience, and how would the customer like this to be improved?
  • Are there solutions in the marketplace that have successfully addressed these issues already in live environments?
  • How quickly can this be implemented, and what initial and ongoing resource will actually be required to make it run successfully?
  • What upheaval would it create within the existing operation? What effect does it have on the customer experience?
  • Are the improvements measurable?
  • Is there enough clean data available to train an AI-Enabled self-service system effectively?
  • Will our infrastructure or existing platform need to be replaced?
  • Is AI definitely the most appropriate way of dealing with this issue?

In theory, a simple chatbot deployment can sit alongside existing channels without integration, but if and when it can’t handle a customer request, the customer experience will be negative as they will have to start their quest again on a different channel. An integrated deployment, ensuring the right connectivity and APIs are in place, can hand over the context and history of the self-service session to a live agent, improving the customer experience and keeping the internal reporting accurate.

It is likely that senior decision-makers within the enterprise have a sketchy or unrealistic expectation of how AI can help within the self-service environment. As such, it is important that the boundaries of the project are clearly understood, with relevant baseline metrics captured before the project, and clear and achievable outcomes signposted so that the eventual level of success of the project can be clearly understood.

Many contact centres may consider a limited, low-risk use case which can be implemented quickly and relatively cheaply in order to demonstrate a quick win and assert the viability of AI within a customer contact operation. For example, increasing the number of self-service interactions through improved AI-Enabled website guidance in certain defined cases is an example of a project which has a clear and easily measured metric which translates directly into call and cost reduction.

Having said this, it is important for contact centres not to sell this to high-level management as being an opportunity to reduce headcount, as it is very unlikely that this will be an appropriate response to the success of an AI-Enabled self-service project, certainly in the short-to-medium term. It may be better for the project to be viewed as improving the customer experience through providing customers with an alternative to a frustrating web browsing experience, ending with an unnecessary and unwanted live call.

While it is important for the initiAI implementation to focus on achieving success within its own terms, it is also important that this is not seen as a tactical point solution with a single end in sight. For example, while the initiAI implementation may be focused on increasing the effectiveness of self-service in a defined area, the longer term view may be to roll out AI into the agent’s sphere, assisting them while on live calls. As such, a roadmap of logically linked business cases can help to establish a long-term vision which can be shared with non-operational senior personnel to help them understand the strategic use of AI across the customer-facing parts of the business.

For example, a simple yet strategic roll-out may look similar to the following:

  • Use a virtual assistant to improve the take-up of knowledge held within the FAQ database, by improving the search mechanism and offering a two-way conversation interface in order to provide more accurate answers. Capture the phrases used by customers in existing human web chat sessions to understand the questions they will ask your chatbot
  • Place this virtual assistant upon the agent desktop in order to provide them with more knowledgeable potential answers within the call
  • Meet customer requests over voice and text through the use of natural language processing, in order to assess customer intent, and provide answers or optimal routing strategies
  • Improve efficiency, consistency and effectiveness of back office processes connected with the contact centre through the use of robotic process automation
  • Deploy analytical AI in order to discover patterns of data relevant to the business that would otherwise not be identified.

Once the process, objectives and outcome are clearly defined, the selection of a vendor and solution can then be approached. In a rapidly growing and heavily hyped market sector such as AI-Enabled self-service, it can be difficult to compare vendors with like-for-like solutions.

For example, in the case of chatbots or virtual agents, on the one hand these can be rulebased, have limited conversational capability and are unable to learn; on the other, they may use natural language processing, engage with customers in order to ask further questions to determine intent, and be capable of self-improvement. The development time, resource and cost associated with each of these types of project are very different, and businesses must decide whether they are looking for a quick win, or whether they have a definite long-term strategy in mind.

Businesses should also consider the type of developer and implementation model that’s most appropriate: some self-service chatbots can be based on off-the-shelf software which is then customised and implemented by an in-house development team, whereas some businesses may prefer to bring in third-party developers with greater experience in implementation. The rate of change within this technology sector is very high, so implementations that are measured in a handful of months rather than longer would seem to make more sense at this point.

At the request for proposal (RFP) stage, businesses may consider asking potential suppliers:

  • What are the current capabilities of your self-service solution and what does your product roadmap look like?
  • How do you propose escalating interactions to live agents if the solution cannot handle it?
  • What metrics do you propose using in order to judge the success of a self-service implementation?
  • What does the timeline of a successful implementation look like? Do you have a reference site?
  • How do you propose to train the AI, and what will our training data need to look like?
  • How do you propose to integrate AI with our existing self-service and live agent support systems, and how much customisation will be needed?

At the initial stage of the implementation process, datasets that any AI models will be learning from must be analysed, cleansed and curated to provide a solid basis for the solution to learn from. Vendors will have dedicated examples of neural networks that work for various business cases such as providing answers to queries or estimating the time taken for a process to be completed. These can be used as a starting point for training the AI model, and to enable it to start making predictions of its own.

While each vendor will have their own framework and architecture, they are likely to follow a similar path involving input, interpretation, action and improvement. Input is gathered by the system - often from a customer - and is then translated into a form which the system can understand (e.g. through speech-to-text or optical character recognition).

Once the data are converted, the AI looks for the customer intent behind the input using NLP as well as other metadata such as location or customer history. Once the intent has been decided with a certain level of confidence, various solutions are considered and presented to the customer. Finally, the loop is closed through gathering feedback about the success or otherwise of the answer, which is then taken into account in future interactions, with the AI learning what works best.

In theory, despite the often onerous effort involved in creating a clean pool of data, the implementation of a virtual assistant or chatbot should not have to materially change the existing IVR or web chat infrastructure, as the AI agent is treated as just another user of this technology.

The more data that the AI has to train on, the more likely it is to succeed. As machine learning works through pattern recognition, this can include metadata and context which may seem somewhat peripheral to the process, but there are likely to be patterns that have not been recognised by human users.

This allows the AI model to understand customer intent and also to be able to discern which customers need to be treated in a way outside of the ordinary (e.g. in an emergency situation, if the customer is likely to defect, or if they have contacted the business multiple times in a short timeframe). Analysing the use of existing data shows the ways in which customers want to express themselves

As with any IT project, testing is key to success but with AI implementations this is even more important. IT systems work on an input-processing-output basis, where the point of the implementation is that the same thing happens every time, reliably and predictably. As the processing element of AI involves elements of learning (and hopefully improvement), the output can change over time.

This does not always end optimally: Microsoft’s Tay bot is an example where Conversational AI displayed inappropriate and offensive responses after interacting with Twitter users. Businesses should be aware that AI solutions, especially in the early stages, may require very close supervision and possibly intervention. Dedicated chatbot testing vendors offer services to make sure the chatbot is working properly before putting it into a live environment.

Measuring the performance and success of an implementation is always vital, but never more so when it is for a highly anticipated and poorly understood solution such as AI. There is likely to be far greater interest in from the higher echelons of the business as is the case for most contact centre technology implementations, and thoroughly understanding the outcome of the initiAI implementation is vital.

There is no baseline set of metrics that every AI-Enabled self-service implementation should be measured against, although in the widest sense, the impact upon customer experience, agent experience and operational change should all be considered. Of course, it also depends on the area of the contact centre business processes that implementation is aimed at improving.

Some examples of AI-related metrics related to self-service include:

  • volume of self-service attempts, segmented by type of interaction
  • customer satisfaction by self-service segment (particularly useful for comparing AI enabled self-service with scripted self-service)
  • proportion of self-service attempts that are handed-off to agents, and following from this, the proportion of these which are handled by a single agent (shows the efficiency and accuracy of routing and the collection of relevant information in the initial self-service session)
  • contact deflection / number of sessions handled entirely by the chatbot
  • length of self-service session (this is related to customer effort)
  • change in inbound call volumes.

END-USER QUESTION #3: WHAT DO SUCCESSFU L AI-ENABLED SELF SERVICE IMPLEMENTATIONS / PROJECTS HAVE IN COMMON? ARE THERE ANY PITFALLS TO AVOID?

Two key components. One is that successful AI-Enabled self-service implementations allow for quick, intuitive configuration and optimisation by business users. This is achieved through a solution that provides an intuitive user interface that allows for updates to be quickly done by the business vs beholden to IT. As a result, the organisation can avoid the pitfall of unnecessarily long cycles internally toward optimisation caused by bottlenecks from too few people being able to impact changes. Once this is broken down, organisations can start adjusting their implementations faster, staying ahead of the competition and proactively defining relevance instead of playing catchup.

The second component is that the solution allows for personalisation Ideally, personalisation is dynamic and based on customer data - who they are, what they are doing, and what they are likely to do next. This means leveraging customer information, customer touch points and contextual data in how the system responds. That translates into not just a more empathetic self-service experience, but also one that is able to manage a larger array of enquiries that don’t have to be escalated to an agent.

AI-ENABLED SELF-SERVICE: AVOIDING PITFALLS

In any technology implementation there will be risks of failure: with AI covering a vast amount of territory and with the potential to be misunderstood by business owners, planning and expectations must be managed very carefully.

  • Expectations of what the implementation can actually achieve must be closely managed. There may be the expectation from senior management that headcount will immediately begin to drop, but in the majority of instances this is not why AI-Enabled self-service is being implemented. Focusing on a tightly defined use case would reduce the risk of implementation delays and expecting too much, too soon. However, it is important not to see even a relatively modest implementation as being a point solution, rather than a single strategic step
  • There are areas of customer interaction where AI-Enabled self-service cannot come close to matching a human agent. Machines simply are incapable of feeling empathy, and even sophisticated sentiment detection at its best comes close to what an ordinary human being can do naturally. Use cases should be focused upon areas where there is a gap in functionality, rather than trying to replace something that isn’t broken, or in areas where even the best human agent could not improve on the AI’s answer
  • AI in the contact centre is relatively new, and with it being so popular, there is a shortage of skills, support and resource within the industry as a whole. In-house technology departments are less likely to have capability, expertise and experience, meaning that the risk of suboptimal deployment and the requirement for third-party assistance may be higher than with other more traditional IT implementations
  • Businesses’ data assets must be in place before implementation of AI-Enabled selfservice, as more complex implementations rely upon having large, clean pools of data that Al can train on and learn from. Without this in place, it will be virtually impossible for any implementation to get close to its potential. The preparation of data will involve having an organised, non-siloed data architecture, a consistent data vocabulary, the means of accessing this data securely and quickly, and the ability to access other pieces of relevant information (e.g. customer-related metadata) in order to include greater context. Without this, it will be difficult for a machine learning process to train itself effectively, or for a chatbot to be able to use all of the relevant data in order to reach a correct conclusion
  • Always have a well-designed and clear path out of AI-Enabled self-service and onto a human agent. Trapping a frustrated customer in a self-service session runs the risk not only of training them not to use self-service again, but also poisons the well for other companies using AI.
  • This is what happened in the early days of email support - customers would try to communicate with one or two businesses via email, and when they didn’t receive a response for days (or ever), they decided that the whole email support channel was unworthy of their time. It took many years to change this perception and to get them to trust the channel again
  • There have been a lot of media scare stories about AI and robots making people unemployed. It is important to emphasise to agents that any implementation is about making their jobs more interesting and effective by allowing AI-Enabled self-service to handle simple and repetitive requests, as well as providing them with more of the information that they need to serve the customer more effectively. While agents are experts on answering customer queries, it may be too much to ask them to spend significant amounts of their time on contact curation as well. As such, businesses should consider how to incentivise power user experts (both inside and outside the enterprise) to help with knowledge management and problem resolution
  • In the AI world, knowledge management is not something that is a part-time job or that can be handled by amateurs. Consider developing more full-time, expert roles to support knowledge bases and to enable understanding of data models and flows across the entire enterprise. AI experts have to understand both data and also the real-life business / customer issues, and this resource can be difficult to find.

END-USER QUESTION #4 : WHAT SORT OF MEASURABLE BENEFITS SHOULD WE EXPECT FROM AI-ENABLED SELF-SERVICE?

Deflection rate and improved customer satisfaction, measured in KPIs like CSAT and NPS. Deflection rate will present a picture of how much the self-service is able to manage on its own. This translates into how much additional scale the organisation can achieve thanks to chatbots. Meanwhile, customer satisfaction metrics give visibility into overall satisfaction of the self-service experience. Both areas can be ones that become benchmarks for the self-service implementation, with goals set to continually improve them through optimisation.

A real-world example of these can be found in how Advia Credit Union has leveraged Genesys DX. Through using the solution, the company has achieved a 75% deflection rate for its 170,000 members with estimated cost savings of more than $400,000 per year through eliminating third party vendors.

THE FUTURE OF AI-ENABLED SELF-SERVICE

Potential uses of AI-Enabled self-service in the customer contact space include:

  • Emails that look as though they have been written by a person rather than a machine, using natural language processing to write content, as well as understand it
  • Tailor information based on the customer’s history and behaviour for marketing as well as service, sending emails at a time when they have been calculated that they are most likely to be opened
  • Increased opportunities for personalisation, as the full customer history can be checked in near real-time, with far more data practically available to the AI than would be for a human agent
  • Machine learning will allow AI to go beyond simply what they have been programmed to do, seeking out new opportunities and delivering service beyond what has simply been asked of them
  • Use of text analytics to assess not only data held within the company, but also in unstructured, third-party environments, such as social media, comments on websites and public forums, in order to learn and deliver proactive service before it is even requested
  • Text analytics can also be used on inbound interactions such as emails, running an AI triage system to assess the priority and urgency of each request in order to handle these more effectively and in an appropriately timely manner
  • Work alongside agents to provide relevant knowledge that may be otherwise take a long time to find, and update the knowledge bases available to humans and AI self-service systems using an automated feedback loop that is constantly improving based on actual outcomes
  • Through understanding multiple customer journeys, Ais will be able to predict the next most-likely action of a customer in a particular situation, and proactively engage with them so as to avoid an unnecessary inbound interaction, providing a higher level of customer experience and reducing cost to serve
  • Speech to text transcription (a key requirement for voicebots) will rise from around 90% to near-perfect accuracy, across a wider variety of accents and speaker types. Improved sentiment and emotion analysis will better identify those customers that need human attention. The level of automation in virtual agents will also increase, as well-integrated feedback loops continually learn and improve responses to standard customer service interactions
  • The ability to join up multiple self-service application technologies to work coherently is key. For example, voice and mobile self-service both exist today as mature standalone solutions that are increasingly utilised, but they typically operate in isolation. Integrating them to provide a joined-up customer experience and journey is vital.

Looking further into the future, businesses’ interactions with customers are becoming a highly polarised mixture of the automated and the personalised. Moving a large proportion of interactions onto self-service works for businesses, and is increasingly popular with a customer base that is becoming more sophisticated and demanding in what it expects from self-service.

We can expect to see self-service using large amounts of sophisticated artificial intelligence, with personal technology applications seeking out the best deals on offer, or interacting with a business on behalf of customers. This leads to the conclusion that many customer-agent interactions will become even more exceptional, such as a complaint, an urgent or complex issue or a technical query that an FAQ or customer community couldn’t solve. It may also be that whole segments of the customer base who don’t want automation at all will still be handled directly by live agents.

Many self-service scenarios suggest a world in which customers speak directly to ‘intelligent’ systems, but the world of the ‘virtual intelligent personal assistant’ (VIPA) turns this idea on its head, postulating an e2e world (in which systems talk to systems), where the customer delegates many of their business interactions to a pseudo-intelligent device.

Storing information on a VIPA device - such as personal preferences, financial details and individuals’ physical profiles - is the first step. Customers of the future will instruct the device to research the best deals for products and services, or to find out the answer to a service query, and to come back to the device’s owner with the right information. The VIPA would ‘call’ the relevant contact centre (which would in fact be either a number of back-office company systems or possibly a live agent in some cases) and even purchase the best deal without having to involve the owner in any way.

The same principle applies to customer service: using the 'Internet of things’ means that, for example, smart appliances such as utilities meters now send their own readings to suppliers on request, and in theory a manufacturer can detect when a part on an appliance is about to fail, and organise a replacement part and engineer visit with the customer’s permission.

VIPAs may be used in association with intelligent agents which roam the web for answers to questions or situations, and could act as a third-party broker between the customer and a business. Price comparison sites act today as a type of first-generation smart assistant, but are entirely reliant on accurate and complete data inputs being provided by suppliers and the site’s owners. If VIPA technology could be relied upon to work, and standards of interoperability between VIPA and businesses were implemented, then this immediate and extensive market knowledge could create a 'perfect market’ for commoditised products and services, with major impacts on the way business operates, creating a true self-service ecosphere.

END-USER QUESTION #5: WHAT WILL CUSTOMER SELF-SERVICE LOOK LIKE IN 5 YEARS’ TIME?

As underlying technologies like NLP improve, automated conversationswill feel increasingly more “human”. Selfservice will become more conversational and will scale to more languages that what is typically offered today. The role of the agent will change as well. Agents will play the role of respected specialist rather than having to be the front line of escalations for self-service gone bad. Proactive and predictive elements will continue to improve and become an expected element. Solutions, such as Genesys Predictive Engagement, already allow admins to build a path, predict segments and outcomes and can sense when there's a need to act. That means automatically triggering a bot, an agent-assisted chat, or a content offer, to drive the optimal next step in their journey. Part of this is done through understanding patterns such as sales abandonments and churn risk so it can trigger a precisely timed action that optimizes your business outcomes and your customer's experience to ultimately increase engagement but also conversions.

ABOUT CONTACTBABEL

ContactBabel is the contact centre industry expert. If you have a question about how the industry works, or where it’s heading, the chances are we have the answer.

We help US and UK contact centres compare themselves to their closest competitors so they can understand what they are doing well, what needs to improve and how they can do this.

The coverage provided by our massive and ongoing primary research projects is matched by our experience analysing the contact centre industry. We understand how technology, people and process best fit together, and how they will work collectively in the future.

e: info@contactbabel.com | w: www.contactbabel.com | t: +44 (0)1434 682244

Free research reports available from www.contactbabel.com (UK and US versions) include:

  • The Inner Circle Guide to Agent Engagement & Empowerment
  • The Inner Circle Guide to AI-Enabled Agent Assistance
  • The Inner Circle Guide to Chatbots, Voicebots & Conversational AI
  • The Inner Circle Guide to Cloud-based Contact Centre Solutions
  • The Inner Circle Guide to Customer Engagement & Personalisation
  • The Inner Circle Guide to Customer Interaction Analytics
  • The Inner Circle Guide to First-Contact Resolution
  • The Inner Circle Guide to Fraud Reduction & PCI Compliance
  • The Inner Circle Guide to Next-Generation Customer Contact
  • The Inner Circle Guide to Omnichannel
  • The Inner Circle Guide to Omnichannel Workforce Optimisation
  • The Inner Circle Guide to Outbound & Call Blending
  • The Inner Circle Guide to Remote & Hybrid Working Contact Centre Solutions
  • The Inner Circle Guide to Self-Service
  • The Inner Circle Guide to the Voice of the Customer
  • The Australia & New Zealand Contact Centre Decision-Makers’ Guide
  • The UK Contact Centre Decision-Makers’ Guide
  • The US Contact Center Decision-Makers’ Guide
  • The UK Customer Experience Decision-Makers’ Guide
  • The US Customer Experience Decision-Makers’ Guide
  • Exceeding UK Customer Expectations
  • Exceeding US Customer Expectations
  • UK Contact Centre Verticals: Communications; Finance; Insurance; Outsourcing; Retail & Distribution; Utilities
  • US Contact Center Verticals: Finance; Insurance; Outsourcing; Retail & Distribution.

Find it useful and want more insights?

Read more
Related content
Webinar

How a Global Bank Turned Complexity into a $60M Opportunity

 Blog Post

Where AI Meets Customer Experience: A Look Ahead to 2025

 Blog Post

AI-Powered Teams: A New Era of Employee Enablement

 eBook

Choosing the Right Contact Centre Platform

 Report

Unlocking the Customer Experience Advantage

 eBook

Blaze Past the Next Customer Experience Frontier with Genesys Cloud

 Video

Reimagining Everyday Interactions with AI

 Webinar

Deliver Outstanding Experiences at Scale with AI Copilots

 Case study

Building a Unified Experience, One Connection at a Time

 Playbook

Win the CX Race with AI-Powered Experience Orchestration

 Video

Generative AI in Action: Unlocking New Dimensions of Customer Experience

 Report

Unlocking the Experience Economy: Insights on Generational Shifts

 Report

Step Into the Future With a Smarter Path to AI Adoption

 Report

Self-Service That Serves Everyone: Balancing Efficiency and Empathy

 Report

AI in CX: Unlocking the Future of Personalised Experiences

 Report

From Vision to Strategy: Crafting Your AI Business Case for Success

 Video

Unlock the Future of CX with AI

 Video

How Experience Orchestration Is Accelerating the Era of AI

 Video

Soaring to New Heights: Virgin Atlantic’s AI-Powered Transformation

 eBook

Stay Ahead: Four Security Megatrends Impacting Your Contact Centre

 Video

Scaling Success: HSBC’s Cloud-Powered Revolution in Customer Experience

 Report

Humans and AI: Shaping the Future of Customer Experience Together

 Report

The Power of Integrated Customer Experience

 eBook

How CX and EX Leaders Are Transforming the Future

 Webpage

Top Customer Experience Trends in 2024 and Beyond

 Webinar

Unveiling 2024’s Top Trends in Customer and Employee Experience