Why AI Is Not a Silver Bullet to Solve Broken Customer Journeys

AI is already reshaping customer service, but Simon Thorpe, Director at Pegasystems, argues the biggest shift is not the technology itself. It is the rise in customer expectations and the pressure on CX teams to deliver consistent outcomes, fast. In this CX Today interview, Thorpe explains why many organisations are treating service as a strategic battleground, tied directly to retention, churn reduction, and trust.

Thorpe warns against the idea that AI is a silver bullet. If businesses layer automation on top of broken journeys and fragmented workflows, they will frustrate customers more quickly, not less. The focus, he says, should be on strong foundations: governance, transparency, auditability, and AI that is connected to real processes so leaders can explain what happened, when, and why. That matters even more in regulated enterprise environments where trust is non-negotiable.

Don’t miss the full interview for a practical take on building trustworthy, outcome-driven AI in CX.

Transcript

SPEAKER_00 0:07

Hello and welcome to CX Today. Businesses are increasingly turning to AI to streamline their customer-service interactions, but consumers still expect seamless effective experiences. I'm Nicole Willing, and to help us explore what works and what doesn't and where AI is headed, I'm joined by Simon Thorpe, Director at Pegger. Simon, it's good to have you with us.

SPEAKER_01 0:28

It's lovely to be here. Thank you for inviting me.

SPEAKER_00 0:31

Great, it's good to have you. So let's start broad. How do you see AI shaping the future of customer service today?

SPEAKER_01 0:38

Well, it is a broad question. I mean, I think it's fair to say AI is changing absolutely everything right now. You know, I think I think that's felt particularly acutely in customer service because you know it's changing what customers expect, it's changing how leaders manage and set up their teams, it's changing how you know we ultimately deliver customer service. Um, and certainly in the conversations I've been having recently, I think most businesses are seeing customer service as a bit of a strategic battleground around service. You know, that I I think we're we're past the days where service is just about handling issues, you know, it's increasingly about retaining customers, reducing churn, building trust, all those sorts of good things. And and I think most people realise that AI is is if not now, it certainly will be table stakes. And a lot of the conversations I'm having at the moment are um, you know, not where and if I should use AI, it's more about how do I use AI safely, how do I do it and use it responsibly in a way that scales, that guarantees an outcome. Um and I'm sure your readers are well versed with hearing about things like agentic AI and you know some of the new capabilities. I mean that that is obviously going to help um uh an awful lot in customer service because let's be honest, customers in the not too distant future will be able to do most things, answer most of their inquiries independently. Um and companies will increasingly increase their ability to anticipate needs, fix problems, you know, all that sort of good stuff. So um so I think you know, AI is gonna change a lot very rapidly, but I think a lot of people are asking the question, where do I start? What are the you know the building blocks to success? Um, and I think the winners are gonna be the ones that architect their uh businesses in a way that will deliver that um uh that guaranteed level of outcome and customer experience that everyone hopes it will deliver.

SPEAKER_00 2:41

Yeah, exactly. Um, and you mentioned, you know, obviously the the rapid pace of change and the questions that it kind of raises. Um, what do you think are the biggest misconceptions that businesses have right now about how they can use AI in customer experience?

SPEAKER_01 2:55

Um I think probably that I mean there's there's certainly downward pressure, understandably, from a lot of board execs saying, look, we need to be using AI because of the vast returns that it promises. You know, reducing the reliance on human labour forces is a big one that we hear a lot of quotes from the analysts about. Um but I I think people need to be careful of thinking AI is a silver bullet, that you just switch it on and it solves all your problems. Um, you know, realistically, um I think AI has a huge amount of potential to change things for the better. Um, but unless it's built around the right foundations, um, you know, it could cause more problems than it's worth. And this I uh idea that switch on AI and suddenly your c your costs fall, um, I think that's a bit of a misnomer. You know, if you're building on wonky foundations, if you're building on wonky uh customer journeys and workflows, you're just likely to upset and frustrate your customers even more, even quicker. Um so I think uh the biggest misconception is how quickly AI is going to get used. Um I think you know, certainly from the conversations I'm having, there's big ambitions, but it's where do you start doing that? What are the use cases, how do you um think about this in a safe way, particularly with big enterprise regulated environments? You know, trust, confidence, governance, all of those factors are really, really important. And I think a lot of people are still toying with the idea where to use it and how to use it. So um and customers are showing signs of of you know they need to be able to trust it as well. Um so um so I think certainly my conversations, a lot of points around operating models, foundational elements, is data in the right place? Are we just layering AI over broken processes or are we actually thinking about real transformation? Because I am seeing you know plenty of examples of experimentation, but turning that from pilot into mass scale, that's taking a bit longer, and I think that is really where a lot of organizations are thinking things through at the moment.

SPEAKER_00 5:09

Yeah, definitely. It does seem to be that challenge, doesn't it, to get from the next stage from the pilot stage to a full-scale implementation. So when it comes to success, what makes AI effective in customer service and then where does it tend to fall short?

SPEAKER_01 5:26

Um I think I mean ultimately AI's got to be able to deliver outcomes, you know, it's got to be able to deliver reliability, predictability. We at Pega talk a lot about predictable AI, um, and it's got to be connected to the business. You know, just you know, layering over AI into pockets and you know, treating it as a bolt-on isn't gonna give the transformational value that companies are after. Um, it's got to be AI grounded in real process. You've got to be able to um audit it, you've got to be able to understand. Um I use this this kind of framework in my brain when I'm talking to clients, and and it talks about four kind of levers that you've got to be thinking about. Um so the first is is governance, you know, you've got to have clear rules, accountability, and you've got to have that baked from the very beginning. Um you then need to think about embedding across the fabric of the business rather than those pockets. I think that's really important. Um, that transparency is so crucial because you know, unless we can tell an auditor or regulator what happened when and why were stumped, um, and that consistency, scalability, you know, are really important. Ideally, we we want to be using what we learn in AI and transferring that across the business. And I often joke with with clients that you know, if you've ever used um a prompt in ChatGPT, if you've ever asked the question, um, you know, think of a question and ask ChatGPT that same question three times, you will get three slightly different answers. Now that might be fine from a creative point of view or a brainstorming point of view, um, but it's not necessarily the reliability that enterprise companies need, you know, where trust is an absolute you know paramount. And we really run the risk, if we're not careful, of um of running too quickly with AI and turning customers off. And let's be honest, you know, we're we're still in a lot of places getting over um the challenges of you know previous implementations, things like chatbots that didn't work, you know, IVR systems that didn't, you know, take customers to the right place. You know, consumers have a healthy level of spec scepticism about this stuff. So we're gonna have to be successful right from the off.

SPEAKER_00 7:46

Yeah, exactly. I mean trust is is so critical, isn't it? So, what do companies need to do to kind of earn customer confidence, you know, in these experiences that are going to be powered by AI?

SPEAKER_01 7:58

Um I think I I think really, I mean, I think there's this this idea that you know customers want to use AI, um, and I think that's growing. I think there's definitely signs, you know, as things like you know, Gemini and um Chat GPT become more part of the consciousness, consumers are becoming more used to it, they're starting to use it in their daily lives. So there is an I suppose a notion of um, you know, if if I can get the answers on ChatGPT, why can't I get the same level of intuitive service when I talk to my banks online um uh online banking support? So there is that sort of customer expectation to a degree, but ultimately it's not the AI that customers want, it's the outcome. They want to be able to get, you know, the reason they're getting in touch, something's gone wrong or there's a problem, or they need help, or then it's value. We need to be able to guarantee that comes um consistently. So I would always start when you're thinking about AI, start with the outcome in mind. Don't start with the channel or um um or you know, necessarily where the data's gonna come from, but think about what you're trying to do for the customer. So those long customer journey maps that we used to know and love in the industry, you know, thinking through that is really important. Um and be clear about you know where you want to add in AI um and where is AI being used and when uh and what can it, you know, can and can't do. You know, we we talk a lot in Pega terms about you know AI not replacing people, you know, it's uh you know using to accelerate and automate, take away the low-value um work, and also really making people shine. Um, and so I think one of the big confidence points and trust drivers um is outcome consistency. Um, you know, there's a real problem of consistency, and it has been for many years in customer service. You know, I go onto the phone versus the website, and it's a different kind of experience. So consistency is going to be really crucial. Um I think you know, being realistic about what AI um can deliver, you know, I think most customers are pretty you know pragmatic. They don't have to um um you know, they don't expect the world, but they expect to know when AI is being used, um, you know, they need to, you know, understand what value it's driving for them. Um and um, you know, being transparent with the end-to-end process is really important, yeah, definitely.

SPEAKER_00 10:33

And then you you mentioned obviously the the role of human agents, um, you know, especially as AI becomes more capable, you have you know multi-agent orchestration coming in. How should businesses kind of approach that, you know, striking the right balance between the AI-driven self-service and the role of the human?

SPEAKER_01 10:52

Um, yeah, it's it's an interesting one. I think uh, you know, I think it's it comes down to a question of how much AI versus how much human, you know, really making that judgment call and being specific about what needs to happen to drive that interaction. Um, so you know, right now a lot of customers would say, if you interviewed them, would say that they get better outcomes from speaking to a human being. Now, you know, that may be right, and one is certainly not going to push everything to automation. You know, there's plenty of use cases where it's gonna be a long way off until we're we're fully automating that because they're you know highly emotional or they carry, you know, a customer really feels the need to speak to a human being about it. So we have to make those judgment calls. Um, but I think AI certainly should be increasingly handling the sort of routine high-volume, low complexity interactions as much as possible. Um the the challenge that I'm seeing in the market is that as as we automate more and do more of that stuff, what I'm not seeing as much of is the shift to supporting the frontline service teams because their work is just getting more complicated, more emotionally driven, high pressure, um, leading lots of complexity. And so we've got to give them the tools to be able to deal with that. You know, they're becoming, you know, they're not just call answerers, they're you know, they're investigators, they're negotiators, they're uh I heard a brilliant term the other day, someone said empathetic problem solvers. And so, of course, they're gonna need some help, you know, that they're gonna need that um that guidance. So I think the future of AI is AI, it's not AI or humans, it's AI and humans, and we've got to think about when that's right to make those judgment calls. And a lot of that comes down to thinking about the end-to-end journey, thinking about the work and how that traverses across the business, and where AI could pick up some of that work and where uh where humans you know really sing would really add value. So thinking about that foundational layer is really important rather than just thinking about AI in sort of pockets of um of places.

SPEAKER_00 13:02

Yeah, definitely. And it's interesting. Um, you mentioned obviously the the agents' role becoming more complex, and I'm thinking, you know, with all the security breaches, and you know, you kind of have like AI agents, you know, f automated fraud almost where you have you know voice cloning and fakes and all that. Are companies thinking enough about the kind of security training that agents are gonna need to deal with this more complex environment that they're in?

SPEAKER_01 13:28

I mean I think some are, absolutely. But I I think we're I think in some ways companies need to be thinking about it faster, um, because you know this technology is moving very, very quickly. I think I was on one of your um uh your podcasts not so long ago talking about how consumers will increasingly start using AI to complain, to you know, get refunds, that you know, that's just gonna drive more work into organizations. As consumers become more savvy, we're gonna have to make sure that our service teams are equally savvy to be able to deal with it and deal with the you know the the AI support that it's giving to consumers and security and and and all of the the the challenges that come with that. So so I think certainly some of the some of uh some of the organizations out there are thinking about this very heavily, but I think there's a bit of a catch-up motion that needs to happen.

SPEAKER_00 14:22

Sure, yeah, that makes sense. Um and you know, you mentioned for the consumer how important transparency is and like the responsible use of AI and disclosure and so on. So um looking ahead, how can um companies deploy AI in ways that feel kind of natural to the customer rather than intrusive?

SPEAKER_01 14:41

Yeah, I think this is going to be the real learning curve. I mean, you know, transparency, you know, it has to be foundational from the beginning, very beginning. Um, you know, it it's not a nice to have. Um customers will see right through it if we try and force them into a way that they're not comfortable with. Um so I you know, I think we're gonna have to work through that trust curve very carefully. Um and we're gonna have to make it very clear that there is a handoff capability to to a human being should they want it, because we're, you know, all consumers are different, and um, you know, we come at these things depending on the use case and the inquiry type on um you know very differently. So I think you know they don't expect perfection customers, but they do expect accountability, consistency as we've talked about, fairness, um, and I think a lot of that comes down to being able to um govern the decision making, you know, give outcomes that are expected, um uh you know, well-defined escalation paths, you know, safe um uh you know data usage, all of that sort of thing. I think like anything, if we can guarantee that consistency and that repeatability, then customers will very quickly adopt. But the problem is they've been burnt a little bit by some of the I don't know, self-service capabilities we've introduced in the market um over the past 20 years, and some of the you know technology that hasn't done what it said it would do on the tin, and and so we're almost having to work twice as hard to make up for past challenges and issues. Um so um, and I I think a lot of this and a lot of the fear that I'm hearing about in the industry is AI is wonderful, it can do so many brilliant things, you know. We've all played around with it, but is it black box? You know, how do we actually um know what it's gonna do, when it's gonna do what it's gonna do, and all that sort of thing. That's why we tie at Pega, we tie our AI um to workflows, because workflows are inherently predictable, you know, they really help govern the you know the business policies and the procedures and the things that you want to work. So um so I always liken it to you wouldn't just sit stick a human being on the phone on day one without you know talking through the manuals. You you can't do this that with an AI. Um so um so you have to make that that that distinction. Um but um yeah, it's gonna be it's gonna be challenging to get through that that uh transparency curve. But I I think if we can show that consistency and repeatability of outcome, customers will adopt very quickly.

SPEAKER_00 17:22

Yeah, sure. No, you've you've given some some really good advice and practical pointers there for people watching. Um I could go on talking about this, but uh that's all we have time for. So thank you so much, Simon, for sharing your insights.

SPEAKER_01 17:36

Thanks for having me, and uh I hope it was useful. And feel free if anyone wants to uh talk further, uh drop me a line anytime.

SPEAKER_00 17:43

Great, thank you so much. So it's clear that you know AI has huge potential in customer service, but getting it right takes thought and design. So to our viewers, um, you know, for more interviews and analysis on these topics, be sure to visit our website, cxtoday.com, and um make sure you subscribe to our newsletter and you can join in the conversation on our LinkedIn community. Thanks for watching, and we'll see you next time.