Is your AI deployment just leading to “faster horses?”

There’s a lot of energy around AI in customer service right now, and rightly so. The potential is clear - faster responses, lower costs and more consistency. In many organisations, early results are already starting to show.

But what if AI is simply helping us do the wrong things more efficiently?

Spend a bit of time looking at the demand hitting most service operations and you start to see a pattern. Typically well over 50% of contact isn’t customers asking for something new, but customers chasing progress, repeating information, or trying to get something resolved that should have been sorted earlier in the journey. The interaction feels like the work, but it’s often just a symptom of something happening elsewhere.

Historically, real people have absorbed a lot of this. They interpret what customers really mean, navigate across teams, chase progress, and apply judgment where the system falls short - acting as human glue to hold together an otherwise fragmented journey. It’s not elegant, but it allows the operation to function.

As AI takes on more of that "interaction layer", that human glue starts to fall away. What’s left is the operating model itself, which is where things begin to get interesting.

Most service models have been designed around the assumption that work is carried out by people. Tasks are split across teams and hand-offs are built in based on some underlying assumptions about human capability (one person couldn't possibly learn all that!) Then controls are layered on top, and performance is managed through targets and quality checks.

It makes sense when you’re co-ordinating large groups of individuals with different skills, capacity and experience. But if some or all of that work is now being handled by an AI agent, some of those assumptions start to look less solid.

Why pass work from one team to another if the same agent can complete both steps? What does “quality assurance” mean when the work is executed consistently every time? How do you think about incentives and performance when there is no individual to manage?

These aren’t just edge cases - they go to the very heart of how the service is designed.

AI won’t just change how tasks are performed, it will start to challenge why the work is structured the way it is in the first place.

So the question isn’t only how much of today’s demand needs to exist, although that’s important. It’s also whether the operating model built to handle that demand still makes sense when the nature of the work changes.

If it doesn’t, there’s a risk that we end up with a very efficient version of a model that was designed for a different world.

And this is where the role of the service leader begins to shift. For a long time, performance has depended on how well people are managed, how targets are set, and how teams are motivated to deliver against them. Those skills still matter, but they are no longer the whole picture.

When more of the work is carried out by AI, the focus moves upstream. Leaders are no longer just managing performance within a system, they are defining how that system works. How demand is understood, how work flows end-to-end, and what outcomes are prioritised all become design decisions rather than operational ones.

That requires a different kind of capability. Less about managing individuals, more about understanding how service behaves as a system, and being able to shape it deliberately. Because once AI is in place, it will follow that design exactly.

Which brings us back to the “faster horses” idea. The risk isn’t that AI doesn’t work. It’s that it works exactly as designed.

We’ve been exploring this with a number of organisations recently, and it’s starting to shift the conversation. Less about where to apply AI, more about how the service itself should be designed when some of the constraints we’ve always worked within begin to fall away.

There's also a real recognition that leadership capability needs to evolve to bridge the gap from "leading people" to "leading AI agents."

If you're thinking about AI, or you're already beginning an implementation, Service Economics can help you review your operating model to ensure you get the best return from your investment, and also help with the leadership development training that will help bring your leaders up to date with the skills they will need tomorrow.