#235: AI Will Transform Contact Centres First

It looks likely that customer service and contact centres will be among the first business functions to be transformed by AI. And not a minute too soon!

Nov 25, 2024

The Only Way is Up

There is an argument that we are currently in the worst age of customer experience. The latest data from ICS (The Institute for Customer Service) shows a record plunge to a nine-year low in customer-satisfaction levels across all sectors of the UK economy.

At the very least, nobody doubts that this is not the golden age of customer experience. The call-centre for the brand NTL when it existed was sometimes referred to as NT-Hell. Anybody who has had to call a call centre for anything - telecoms, banking, or retail, knows the challenge of dealing with contact centres. All too often you encounter people who can’t solve your problem, have to follow a script, and perhaps are not really incentivised to help you resolve your your real problem, focusing instead on superficially dealing with your call.

“The latest UK Customer Satisfaction Index (UKCSI), which since 2008 has benchmarked customer satisfaction at over 275 of the UK’s leading organisations, is 75.8 (out of 100), a drop of 0.8 points compared to July 2023 – and 2.6 points below its highpoint of 78.4 in July 2022. The July 2024 UKSCI now sits at its lowest level since July 2010” - The Institute of Customer Service, 2024.

Adding Up The Moments of Truth

In 1981, when Jan Carlzon was appointed as the CEO of Scandinavian Airlines, he faced an industry that was increasingly competitive thanks to deregulation, and declining profitability. He saw the opportunity to drive transformation through customer experience and empowerment of frontline employees. Carlzon had already demonstrated customer service success with Linjeflyg - Sweden’s domestic airline, and doubled down on his principles. He proceeded to redefine SAS and the airline industry with his core idea of the ‘Moments of Truth’ which he defined as every interaction that an organisation has with its customers, and which is the title of his influential book from 1987.

While not a one-size fits all solution, Carlzon demonstrated that delivering great customer services was a cornerstone of organisational performance. He isn’t the only one though. Although this is an intuitively obvious point, let’s look at some findings. According to Bain & Co, an increase in customer retention of 5% increases profits by 25% - 95%. According to a study by Wharton School, referred customers are at least 16% more valuable over their lifetime than customers acquired through advertising. American Express found that 70% of customers were willing to spend more money with firms that offered better customer experience. Deloitte report that customer centric companies are 60% more profitable than others. We all know how Amazon’s continued prioritisation on customer experience has taken them to unprecedented success. Zappos, the online shoe retailer, was another firm obsessed with customer experience and was aquired by Amazon in 2009 for $850m which was over 75x it’s previous years EBITDA.

So Why are things so terrible?

With this overwhelming evidence for the powerful and positive impact of customer service on financial performance, why is customer service so bad across the board in 2024?

One obvious answer is that digital self service with automated tools and systems has been steadily improving, and has often motivated companies to err on the side of technology, and over-ambition. The dramatic difference between digital self service and a human interaction based service is another factor here. Human based interactions can be 5x-10x more expensive. This is why on service like Uber you will really struggle to get to a phone number to call. The reality is that the digital channel can address 90% of queries, but that comes at the cost of the 10% who really need to speak to somebody. I found out the challenge last year when I was helping my sister trace a phone she had left in an Uber taxi (it was found and delivered but it was a difficult process).

A more likely answer though, is that driving up financial value through customer care is hard work, requires a lot of effort and intent across the business, and takes time. Conversely, the temptation to outsource customer experience with an underlying agreement to save costs is fuelled by the immediate and relatively easier impact on the bottom line. The paradox here is that over time it is likely to negatively impact the financial performance, but planning and reporting horizons may be more short term (quarterly) oriented. You might remember that Jeff Bezos famously ignored the analyst feedback and skepticism for years, instead opting to build on customer satisfaction. Not everybody has the vision to play that long game.

Does That Mean Contact Centres Are Bad?

As with many other areas of business, the intent and incentive structure here probably matters more than whether you outsource or ‘insource’ your customer contact. If the focus is primarily on reducing the cost of customer care then both internal and external options will lead to a race to the bottom, and customer service will in all probability, suffer. If the primary intent is to deliver great customer service with cost as a non-overriding consideration, then either insourcing or outsourcing will drive customer value. My personal example is from Sky TV. By any accounts this is an expensive service, and I’ve been contributing to the Sky coffers for 20 years now. But when we spoke with Sky about moving to fibre, the external junction box was replaced the same afternoon (by BT) and when the Sky engineer came home at the pre-agreed date and time, he inspected our property, took his measurements and connected everything in 90 mins that he was there for. I’m sure Sky have their detractors, but in general over the years, I’ve found the Sky engineers to be incentivised to resolve any problem and satisfy the customer.

So the current malaise of the contact centres and plummeting customer service scores may be more to do with the financial motivations, incentive structures, and the low priority assigned to customer care rather than the fault of the contact centre industry. Having said that, thanks to AI, all of this is likely to change.

Digital Time

With the advent of digital systems, companies have an additional lever to pull. Well designed self service on websites and apps can take care of a significant percentage of customer queries and problems. It took me half a dozen clicks and less than 2 minutes this week to cancel my credit card on the app and the new card is already sitting on my kitchen counter waiting to be activated. But here too, you can have too much of a good thing. Intoxicated by the opportunity of self service, many companies have simply made it next to impossible to contact them over phone, even though the self service doesn’t solve your problem.

For consumers, though, the digital world has changed the perception of time. Thanks to mobile phones and digital systems, you can achieve a lot in a few minutes - be it at the bus stop, or at home. So the minutes have become more valuable. Whereas spending 20 minutes to get something done a few years ago might have been okay, today it’s annoyance value is much higher, measured by all the things you could have done in that time. This combination of companies choosing cost over your time, and time itself becoming more precious, is probably a reason why dissatisfaction levels with contact centres and customer service is boiling over.

Enter AI: All Change

I’m sure you’ve heard or read about Klarna (the Buy-Now-Pay-Later Fintech) and how they are implementing AI for customer service. The Klarna AI was reportedly doing the work of 700 people, and had 23m chats when the news broke 6 months ago. The numbers are likely to have improved since then. The AI is more accurate, so repeat calls have come down by 25% and it takes 2 minutes compared to 11 for clients to complete transaction. And it had an estimated $40m impact on Klarna’s bottom line. The Klarna example alone suggests that AI is here to revolutionise the contact centre, and to a broader extent to change customer service itself. We might be on the verge of this great explosion of AI in customer care. Aside from software development, customer service is likely to be the first commercially transformed area in age of AI. To use a weather metaphor, it’s where the AI hurricane makes landfall on the shores of business processes.

How will this play out? Potentially, in 3 stages. At TCS we use a model of Assist, Augment, & Transform. This is a good way to frame the evolution.

In Stage 1, aka assist, we will largely see AI supporting humans. This means that rather than directly exposing AI chatbots to all customers and problems, organisations will enable AI behind the scenes, so the human agents can access AI to better address customer queries. Also in stage one some of the simpler and repeatable issues will get addressed by AI, which will in turn allow humans to focus on the more complex problems. The next 12-18 months will see a significant roll out of this kind of AI. 80% organisations will have introduced this kind of AI into their contact centres. In stage one you would expect to see a measurable bump on key KPIs such as call handling time, first time resolution, and so on.

For example, consider this extract from an article about AI’s impact on contact centres in Philipnes: “Bernie now uses ChatGPT and Bing to compile all the technical information he needs for a query in less than five minutes. It’s doubled the number of customer complaints he can handle in a day. Yet, his workplace — a U.S.-based, publicly listed “customer experience” giant with employees across more than 40 countries — forbids its agents from using generative AI, according to Bernie.” It’s a reminder that even stage 1 solutions are being embraced by individuals before organisations, in some cases.

In Stage 2, aka Augment, we will flip the model and have AI picking up the majority of the challenges and humans supporting the AI - for governance, specific queries, and training. Over the next 3 years, it’s quite likely that 80% organisations will have reached this stage, or actively implementing for this model. In stage 2, we will see dramatic shifts in some of the KPIs. The Klarna example of 11 to 2 minutes is a good example. In this stage, businesses will get the additional benefit of being able to handle spikes and troughs with a more flexible cost structure (low fixed costs), as you can in principle spin up more instances of agents as required.

It’s probably worth mentioning here the obvious point that these changes are unevenly distributed. You might argue that Klarna is already at Stage 2. But bear in mind that Klarna has a purely digital product, is a digital native platform so it has less legacy systems, and has a very specific and narrow product offering. All of these probably help it move faster to stage 2. A typical large telecom company has dozens of different products, many layers of legacy systems, and hundreds of varieties of customer queries. And it sometimes has to send an engineer into a customer’s home to fix a router, and/or update a legacy service application. For such a business getting to stage 2 will obviously take longer.

In the 3rd stage - transformation - we start to change the fundamental assumptions underpinning the first two stages. For example we could see a scenario where 80% of the problems are solved before clients call. So the client simply gets an update to let them know a problem was identified and solved. The KPI shifts from the number of calls addressed to the number of calls eliminated. In that world, the customer contacts may be more outbound calls to communicate good news and create upsell opportunities. This could mean that rather than reducing the mean time per call, we’re looking to maximise the time per call - to actually drive engagement.

Advantage AI

Beyond the story of the 3-stage transformation, there are 3 more ways in which AI will improve customer experience at Contact Centres.

The first involves time - the average employment term of a contact centre employee is under two years, which means that every 2 years you have to replace, rehire, and retrain. Think of how this might work over 5 years. You would have reached your third generation of contact centre employees, each one requiring fresh training and often starting from scratch. On the other hand starting at the same point, your AI would have gotten exponentially better over the same 5 year period. I believe that in 5 years from now, this won’t even be a debate worth having, such will be the gap between AI based and human based customer contacts.

Second, all of the above is true even assuming that there is no further improvement in AI capability as we know it. But this is not true. AI tools are being created with blinding speed every other week. Areas where AI will improve significantly include empathy, voice interactions, languages, and accents. It feels like AI capability overall is blasting through Moore’s law - doubling in capability every 6 months. Discussions around AGI are becoming more real. According to this piece in the Economist, Sentient Machines, run by Danica Damljanovic, formerly of Apple, can identify “acoustic features” in spoken language, including hesitation, uncertainty, sadness, even sarcasm. There are plenty of studies now pointing to the improved performance of AI in empathetic responses, from healthcare, to standard emotional awareness tests. So the 5 year picture might tilt even more starkly in AI’s favour.

Third, Root cause analysis and training will get better, our ability to work with AI will improve, and every interaction will generate data that can be better used to train AI but also to improve internal processes. So the use of AI and human intelligence to actually rewire processes, structures, and incentives in businesses, may actually have a huge impact on the incidence of the problems we face today. The idea of hyper-personalisation - i.e. being able to use all the available data available for a customer to truly personalise every interaction is also obviously something an AI system can do better than a human being. Let me give you a live example.

I have a debit card from a bank that clearly values my business. The card came packed in a manner that would have not been out of place for a special birthday present. The ornate box, the layers of wrapping, the effusive messaging, and the bundle of services that come alongside the card all suggest that this is a premium product. But I couldn’t activate the card. Do you know why? Because they had a digit missing in my mobile number. And not just any number, it was a number in the country code. And I found this out after we (my wife, but I use the royal plural) spent hours over multiple calls over many weeks to the call centre to find out why. This kind of root cause analysis for why a card is not getting activated will be childs play for an AI system if it’s allowed to analyse the system. The problem was identified 3 days ago but hasn’t been solved yet. This is actually a data validation problem and shouldn’t need AI to resolve at source. But retroactively, AI agents can be designed to run a series of cross checks on existing data to capture anomalies (between address and country code, for example) to find and eliminate errors across all customer and product data. With an AI system you won’t have to specify which data sets to compare. It can be designed to discover inconsistencies across a range of data sets. We recently helped one of our retail clients to do something like this, through our Pace Innovation work.

Share Innovation Exchange

The Range of AI Capabilities

The change in customer care won’t happen as a single incident. I’ve already explained the 3 stages above. But the trajectory is clear - AI gets smarter, we will see it capture an increasing part of customer service work. (See my earlier post on this)

If we were to consider the call centre operator as an arbitrary level of capability that Gen AI is striving to achieve, we could ask what the current gap is and by when that gap will be crossed, if ever.

The TCS CEO K. “Kriti” Kritivasan said recently that the future of call centre is no call centre at all. After all, with AI we should be able to solve problems almost before the client knows there’s a problem. So the bulk of calls should be messages to the clients informing them of problems being solved. Of course clients will always have questions, or clarifications, and the purely predictive model could be years away. But it would seem from the Klarna example that AI’s inroads into the contact centre operations have already started.

For any role, there is a range rather than a single point of complexity and capability required. For the specific instance of contact centre roles, the picture might look like this.

You could extend this model to any job - and consider the range of activities and capabilities required. But for a number of reasons, customer service and call centre work very likely be an early impact area. Venture Capital firms certainly seem to think so by the volume of investment in AI based customer service start ups.

The Human Impact

As with any technology driven transformation there will be a human impact. Over 800,000 people in Britain work in contact centres. Nearly 3m people do so in the United States. And of the 17m global contact centre employees. 1.5m are in the Philippines, with a similar number in India. The macro picture is that as with previous generations of tech, it’s probable that AI will create as many new jobs as it replaces. But at a micro level, the people who lose these jobs will not typically be able to do the new jobs without a significant skill upgrade.

For example Telstra found that AI tools had resulted in 20% fewer follow-up requests from customers, which allowed it to roll out the tech to all call centers, leading to the removal of 2,800 jobs across its operations.

This will require organisations to approach this transition with a human lens as well as financial efficiency, as the global scale and speed of this transition will likely beyond anything we’ve seen before. Countries such as Philippines which has a high dependence on the contact centre industry are already putting in place measures to manage this transition. Some estimates put the number of jobs impacted over the next 5 years at the 30,000 mark, though even this could be conservative given the pace of improvement of the technology.

The Upside

As an organisation, though, the real upside here is the ability to truly transform customer experience and take it to a whole new level. Those approaching this as a way of cutting costs will achieve exactly that. But some of our more forward thinking businesses I know are seeing a simultaneous improvement in client satisfaction metrics, operational improvement, as well as cost savings. And this is the start of the bigger AI transformation - so the quantum of learning opportunity here is also huge.

Personally I can’t wait to engage with an intelligent and well designed AI that can understand and resolve my problems efficiently and in a way that makes the most sense for me.

Events

CoStar Launch

I recently attended the launch of the CoStar Lab - this is an incredibly exciting initiative and for me a glimpse into the way film technology has evolved - a robotics meets mechanical engineering - meets algorithm - meets screen-tech - meets imagination for scenes such as we say in films like Gravity.

BSI Innovation

I also attended the BSI Innovation Management event and was reminded of the maturing of innovation as a function and the ongoing evolution of innovation as a profession and practice.

The TCS Pace AI Event

Last week at our Pace AI Event, we had clients, partners and lively discussions about whether whether we are tinkering or transforming with AI. With a keynote and panels, the answer was I think we need to tinker AND transform. Practical lessons were shared from our clients who have been running AI programs. And it was interesting to see Google, Microsoft, AWS, nVidia and Anthropic on the same panel talking about the future of AI over the next year.

thanks for reading and see you soon.

Innovation Exchange

#235: AI Will Transform Contact Centres First

It looks likely that customer service and contact centres will be among the first business functions to be transformed by AI. And not a minute too soon!

More AI Reading:

Other Reading

Events

Discussion about this post