Keeping It Human in a World of AI

In this episode of CX Now, Lauren Gold sits down with Tue Søttrup, CEO of SmartRole. Tue's career in CX began 20 years ago on the front lines as an agent himself, and has since spanned a mobile phone company, a bank, an online bookstore, and seven years building customer service software at Dixa — before founding SmartRole, a training platform for customer service agents.

This interview has been edited for clarity.

Lauren Gold: One of the things I was so excited to hear your perspective on — and we hear it a bit in the industry — why do you think CX is so ripe for the AI transformation? Why do we have the opportunity to raise the bar and set the standard for what AI transformation looks like?

Tue Søttrup: I must counter that and say that's not what I'm seeing in reality. I think that's the general sentiment, but we're pretty far away from that in reality. I think a lot of organizations want to use AI to solve a headcount problem, and they use it to have fewer people, so they build chatbots and voicebots, but it doesn't really work yet. It will, but it's going to take longer than what we think.

But the positive thing about it is that it does allow organizations to help customers do things faster and easier than they could before. And that frees up time for agents to focus on more important tasks and more complex, high-stakes conversations.

Lauren Gold: A fresh hot take — I appreciate that. And it makes sense. Definitely that notion of the boardroom mandate versus what's actually happening out on the ground. So when we think about how AI is elevating our current humans, what does keeping it human mean to you in a world where AI is becoming more and more prevalent?

Tue Søttrup: I think a good example of where AI can be really beneficial and create more human connection is when you look at a long conversation and then you have a summary. As an agent, when I get a conversation, there can be a summary. I don't have to read back and forth to understand what it's about. AI is really good at digesting that information and presenting it to you. So I can start a conversation saying, "Hi Lauren, thank you for contacting us. I can see you had an issue with the delivery." And that is of course an improvement because you don't even need to state what the issue is about because I have that information. So that's one area where you can deliver a more personalized experience that resonates with what the customer's expectations are.

Lauren Gold: Yeah, I'm laughing only because we have an amazing product that actually does just that. The way we think about it is kind of like your A players hire A-plus players — the A player agents are going to look to those AI summaries to help supercharge their day-to-day and allow them to 10x themselves, be more efficient, be more customized in their approach with their customers. That's why I was smiling as you were describing it.

Do you find that the agents you're working with are excited about those additions, or how do you think about the folks that are maybe still a little hesitant to adopt what you just described?

Tue Søttrup: I think they are a little bit ambivalent about that technology because on one hand they are afraid it will mean they will lose their job. On the other side, there are certain things that make their job easier. With the summaries, there can be translation, there can be suggested answers. A lot of us are using AI to draft answers, and they've reached a level where they can draft it in our voice — even within platforms like yours, where it will learn from the agent's previous answers and write something in their tone. You can write some keywords and it will do some of the heavy lifting. And in some cases, AI is better at writing than humans are, and they also make fewer spelling mistakes.

I think that's where we saw AI really start to take off in a solution like Grammarly, which helped us write with fewer mistakes and more clarity. That makes it easier for the recipient to understand what you actually want to convey to them. And for me, it means that what I'm trying to say is more correct than if I don't have that help. That's where I see AI coming in in many situations — you find a small problem and you use AI to make that problem go away. With Grammarly, it's spelling and clarity. With summarization, it's not having to read long conversations. And when I need to respond to a customer, you can also look at previous conversations — what answers worked, what delivered a high CSAT — and that means I'm guided to a solution faster.

Lauren Gold: Yeah, that's great. I'm smiling because I'm humbly reminded of all of those great AI assistants that are helping me get through my day and be more concise or have better grammar and spelling. I love having my army of AI agents helping me.

Tell me, Tue — when you see companies, and I know you're out in the industry speaking with quite a few — what do you see companies most commonly doing wrong when they rush to automate a customer experience? Are there any common threads you see that are consistently going in the wrong direction?

Tue Søttrup: Yes, and for me it's really simple what's going wrong. It's exactly what you stated — they bring AI to solve a problem, and they need to go the other way around and look at the problem they want to solve and see if it's something that can be solved with AI. In general, we see a lot of enterprise AI implementations that are failing. I think we're talking 80 to 90 percent that actually fail. And it's because they are going at it from the wrong angle. It's a mandate from leadership to reduce headcount. And if you just try to take AI and force it onto a problem, that's going to create a ripple effect of other problems.

But if you start out by finding the things agents are doing on a daily basis that can be automated with AI, then agents can focus on something else.

Because when all the easy tickets are handled, you can look at the rest of the tickets as escalations. The customer already spoke to a chatbot or a voicebot, it couldn't solve it, and now they want to speak to somebody who can help them. So their expectations have increased significantly. And that puts more pressure on the agents. It also means you need to train them in a different way, because everything will be an escalation — there are no easy tickets to learn from and no mental breaks. Agents like the easy tickets. "When will my order be delivered?" — once in a while. If you just have the hard tickets, it's a very tough day.

Lauren Gold: That's such a great point — how they go about their day, closing tickets, driving great CSAT, increased resolution time, decreased handle time. When every ticket or conversation is complex and requires empathy, understanding, and careful tone, there are no easy wins or a way to generate that momentum. You really need to change your mindset throughout your day as to how you approach it. In some regards it's great — you're spending your time on more interesting and complex issues — but like you said, it's hard to get through the day without some of those easier wins giving you some wind behind your sails. So that's a call to be mindful of.

What is the signal CX leaders need to look out for that tells them their AI has maybe gone too far — when AI is crossing a line in that human interaction?

Tue Søttrup: It's about looking at the guardrails you need to put in place. We've all seen some horror stories. I won't mention them here, but they say things that are wrong. They can be very convincing and can make up terms and conditions, make up airfare rules and things like that that are not aligned with actual policy. So you need to ensure that what you're doing is better than what you did yesterday. When you implement something with AI, what does it allow either a customer or an agent to do tomorrow that they couldn't do yesterday? And what's the value of that, especially when it comes to the customers? Because a lot of organizations have a tendency to implement this for the sake of implementing it — the mandate from up high — and sometimes they forget what problem they're actually trying to solve.

If we look at something like chatbots, that technology actually goes way back, more than 20 years. And we still have pretty bad experiences with chatbots — we must admit that. And now we are trying to put voicebots on top of it. A voicebot is essentially a chatbot that can speak and understand what you're saying. But if the underlying foundation is still just a chatbot that wasn't good, you're not going to make it better.

So you have to start out looking at how you can design a better conversational experience for customers, where they find the right answers. And also identify the customer context where you know this is not something AI should solve — because in 80 percent of cases it won't be able to. Identify those right away and send them to a human. Get the humans to do what they're best at, and get the AI to do what it's best at.

Lauren Gold: Yeah, and that reminded me — we do something here called our Thought Leaders Community, where we bring CX leaders together. About nine months ago, we had a conversation with a lot of CX leaders, and everyone knew they needed to do AI. It wasn't a question of if, it was a question of how. So many, especially in the retail space, were sharing that they have to overcome years of bad chatbot experiences and help bring their customers on this journey with them. They had almost visceral reactions — we don't need our customers to think we're going back in time, we need to bring them forward. How do you do that change management, not just with your own team and agents, but with the actual consumers experiencing it at the end of the day?

And as you said, where does AI keep the human in the loop and make sure you can pass things back and forth throughout an interaction — where AI can help amplify the agent, but when something becomes really critical, complex, or sensitive, it knows to hand it back to a human agent to give it the finishing touches.

Okay, so Tue — how should frontline agents be trained to work alongside AI without feeling threatened by it, or even deferring to it too much?

Tue Søttrup: That's something that's really important and I see two avenues that need to be taken. Number one is to include them in the discussions and also ask them how they can contribute to make this a better experience. They see the other side of the picture — they talk to customers every single day. You can ask them, "What are the kinds of conversations you're handling that you feel are wasteful for the customer and for you?" Get that insight from them and think about how that can be implemented in a way where those conversations are handled by AI and they can focus on something more valuable.

Because they are concerned about losing their job. They are concerned about what impact it will have. But the more you talk with them about it, the more you share about the strategy, and the more you include them in the design of the future conversations with AI, the better it will be for them.

Lauren Gold: Yeah, I think that's great — leading with transparency, being really open, but also open to feedback and helping them shape the transformation and define some of those AI use cases. That's really important.

How about this: what does great AI-assisted empathy actually look like in a live customer conversation? We've referenced the notion of human in the loop and AI guardrails, but how do you build a culture where your team sees AI as a tool that elevates their work rather than a replacement for it?

Tue Søttrup: It's something that should assist them. Just like the rest of us, we are very susceptible to adopting things that make our lives easier. If that's a summary, an automatic translation, a suggested answer, or a recommended support article, I'm going to take it because it makes my job easier. There is a lot of AI functionality that agents want to use all the time — spelling correction and things like that. It's really easy because they see: I achieved my goals faster using this technology.

So include them, understand what's valuable for them, measure the adoption and usage, and focus on the things that actually create value. They will use it because it makes their job easier. If they can see that they use this tool and they get a higher CSAT and handle more customer contacts per hour, they will use it.

Lauren Gold: Yeah, I love that. And even celebrating those wins together as a company — not just at the individual agent level. Everyone is cautiously optimistic about how they can go on their own personal AI journey. People are at such different spots — laggards, major innovators, trailblazers. Celebrating those wins, and sometimes we use the expression "building out loud" — making others aware of what you're doing and why and how you're experimenting — I think makes it a little more transparent and open and easy as we're all learning alongside each other.

I'm going to put you on the spot a little bit. What is a customer experience — one you recently had as a consumer — that reminded you why the human element is still here and still matters, even in this age of AI?

Tue Søttrup: I think we've all had the experience where we started out on a website trying to find a solution, then spoke to a chatbot, maybe a voicebot, and didn't get the help we needed. And then you're transferred to a person. That's usually how a lot of these conversations end up. And when you then meet a person who feels engaged, understands your problem, and is able to solve it, you end up having a good experience.

It's kind of the service recovery paradox — you remember how you felt after a problem was solved rather than the problem itself. So you have an opportunity as an organization to correct something that went wrong when you create a good experience. That's about guiding the customer to find the answer in the right place. And if they can't find it, there needs to be a back door — a way out so they can get to a human fast and easy. And that human needs to be equipped with the knowledge, understanding, and empathy to solve the issue.

A lot of organizations really need to focus on transparency in what they're doing. It has to be transparent that you're speaking to a bot and these are the limitations for the bot. And it has to be distinctly different from what a human can deliver. Humans should be able to think outside the box, be more empathetic, understand the problem, and come up with the right solution. The AI can't do that because it's built on an algorithm. So there's an opportunity for organizations to really upskill their human agents, so you feel a different experience when you engage with a human rather than a chatbot.

Lauren Gold: Yeah, and like you said, I think we can all very quickly point to some very positive experiences and then maybe some others where you're just trying to get a human back on the line and get some answers quickly.

Okay, this is a bit of a spicy one. If you could get every CX leader in a room to stop doing one thing tomorrow when it comes to AI, what would it be?

Tue Søttrup: Trying to solve everything with it. One good example — there was a report from Gartner that came out a couple of weeks ago stating that 50 percent of the organizations that are firing customer service agents because of AI will hire them back by 2027. There is a gap between what organizations think AI will be able to do and what it can do in reality. There was also a big payment provider that let 700 agents go and then turned around and said, it didn't quite work, so we're hiring some of them again.

Organizations are thinking they can solve a lot of things, and CX leaders are part of that journey. They need to look at what problems they can actually solve with AI and really take a step back and see it from the customer's perspective, because a lot of these implementations are done from the inside out — to reduce headcount and cost. That brings, I would say, negative results. But if you start with the customer journey, take a walk in the customer's shoes, buy your own product or service, and see what part of that journey could be better — and then start to implement that.

Lauren Gold: Yeah, I think that's a great call to have that perspective and a long-term view as you go off on that journey.

Okay, our last question today: five years from now, what does the relationship between human agents and AI look like in the best customer experience organizations?

Tue Søttrup: It's a really good question. I'm going to look at it with a positive perspective and talk about what I hope.

I hope that when I have an issue, I can contact the business and maybe I can get my own AI agent to do it — and they can communicate with another AI agent. That could be a billing issue, a delivery issue. And then without doing a lot of work from my side, I'll get an update on that state.

I also hope that a lot of organizations will implement AI to be proactive instead of reactive. There is so much information to be gleaned from deviations in delivery, for instance, or in different flows — and acting upon it, keeping the customer informed. Just saying, "We know that your order wasn't delivered on time. We're looking into it and we'll get back to you with an update soon." That doesn't happen in a lot of organizations today. And that's also an area where AI can benefit.

And when it comes to the human aspect, I see customer service agents transitioning into roles as specialists, solution consultants, trusted advisors — experts in everything the bot can't handle. That will require a higher level of skill, a higher level of understanding of customer needs, and a higher level of critical thinking and problem solving. We will probably see fewer customer service agents, but conversations will be longer. They will be more complex. But they will also be the conversations that create value and loyalty between brands and customers.

Lauren Gold: Wow — I can't think of a better note to end on: value, brand, and loyalty in the customer experience. That really ties it up with a bow. Tue, CEO of SmartRole — this has been so insightful. I really appreciate the conversation and the insights you brought today.

Tue Søttrup: Thank you, Lauren.