The chatbot did exactly what it was designed to do
The chatbot did exactly what it was designed to do. That was the problem.
A customer wanted to cancel an insurance policy because their partner had passed away. The process was simple. The intent was clear. The workflow was available. So the system helped. It asked the right questions, collected the right information and confirmed the cancellation. It probably saved time for the customer service team.
Nothing broke. And still, something was wrong.
Because this was not just a cancellation job to be done.
It was a person experiencing a moment of loss.
That is where many AI conversations start to become uncomfortable. Not because the technology is not good enough, but because the organization never made a real decision about where automation should stop.
Most AI projects begin with a practical question: can we automate this? It sounds innocent. Sometimes it is even useful. Customers do not want to wait for simple answers. Teams do not want to repeat the same manual tasks forever. Organizations need to work faster, cleaner and more consistently.
But somewhere along the way, that question becomes too small. Because if the only question is whether something can be automated, the answer will often be yes. The better question is harder: should this moment be automated?
That question forces a different conversation. Not about tooling. Not about efficiency. But about judgment.
For years, customer experience has been treated as something we improve. We improve a journey, a page, a touchpoint or a message. We reduce friction, remove steps, speed things up and personalize the next interaction. That work still matters, but AI changes the weight of those decisions. A bad button affects one screen. A bad automated decision can shape thousands of customer moments before anyone notices.
AI is exposing decisions that were always there
That is why AI does not just accelerate customer experience. It exposes how customer decisions move through the organization.
Who decides when a chatbot is enough? Who decides when a human should step in? Who decides which signals count as emotional, risky, sensitive or simply too important to automate?
In many organizations, those decisions are hidden inside workflows, rules, prompts, tickets and backlog items. They are not treated as strategic decisions. They are treated as setup. Until the customer feels the consequence.
The uncomfortable truth is that AI is not creating these decisions. It is simply making them visible. Every automated journey contains assumptions about speed, cost, effort, risk and empathy. Those assumptions existed before AI. They were just harder to see. Now they sit directly in front of the customer.
This is the shift I keep seeing across organizations. They are not short on data. They are not short on tools. They are not even short on ideas. They are short on decision systems.
Over the past decade, companies invested heavily in collecting customer feedback. They built Voice of Customer programs. They mapped customer journeys. They implemented analytics platforms. They introduced experimentation programs. For a while, the challenge was finding insights.
Today, the challenge is something else.
How do those insights actually influence decisions?
Feedback is analyzed, but does not always reach the teams that shape the experience. Insights are collected, but do not always change priorities. AI is introduced, but the deeper question stays unanswered: what kind of customer decisions do we want this organization to make?
That is not a technology question. It is an operating model question.
Five questions every leadership team should answer
Before scaling AI in customer interactions, every leadership team should be able to answer five simple questions:
- Which customer moments should always allow a human handover?
- Which signals tell us that a situation is emotional, sensitive or high-risk?
- Who owns the decision rules behind automation, beyond the technology team?
- How do we measure whether automation improves trust, not just speed?
- Where should AI support employees instead of replacing the conversation?
These questions look simple. In reality, they force organizations to define the boundaries of automation before customers discover them on their own. They move AI from a tooling discussion to a decision-making discussion. They force leaders to think about responsibility before they think about efficiency.
Capability is not the same as care
The insurance example is simple because almost everyone feels it. A chatbot might be technically capable of handling the request. But capability is not the same as care.
There are moments where speed is kindness. A delivery update can be automated. A password reset should be automated. A product recommendation can be automated.
But there are also moments where speed feels like distance.
Grief. Fear. Vulnerability. Confusion. High financial impact. These situations ask for something different. Not necessarily a human from the very beginning, but at least a system that knows when the nature of the interaction has changed.
That is the real maturity challenge. Not whether an organization uses AI, but whether it knows what kind of situations deserve human judgment.
For years we measured customer experience through journeys, channels and touchpoints. Increasingly, I believe we need to start measuring it through decisions. Not because journeys have become irrelevant, but because every journey is ultimately a collection of decisions. AI simply forces us to confront that reality sooner.
The discussion becomes even more important as AI evolves from automation toward increasingly human-like interactions. As I argued in Synthetic Empathy, the challenge is no longer whether AI can mimic human behavior. The challenge is deciding when it should.
The future belongs to organizations with boundaries
The companies that get this right will not be the ones with the most automation. They will be the ones with the clearest boundaries. They will know where AI helps, where AI should inspire, where AI should support an employee instead of replacing the conversation, and where human judgment creates more value than efficiency ever could.
The future of customer experience will not be defined by organizations that automate the most. It will be defined by organizations that understand where human judgment still matters.
Because trust is not created when everything works.
Trust is created when customers feel that someone knew when not to automate.
That is why the most important AI decision is not what to automate for customers.
It is what you don’t.
This article builds on a conversation I had with Ben Foden on the CX Heroes podcast after being nominated as a CX Leader to Follow in 2026. One question from that conversation stayed with me: if everything can be automated, where should automation stop?
Watch the episode here: CX Heroes #50 – Tim Thijsse.