I read an article today that used the phrase “intelligent experience” to describe how AI is transforming customer interactions. The phrase stuck with me—not because I disagree that AI is powering some intelligent experiences, but because of who gets to decide what counts as one.
I may be a hard grader, but most of the AI bots I’ve engaged with—voice, mobile, and web—have fallen short of intelligent. There is “intelligence” at work in the background: language models replacing static decision trees, smarter routing, more natural phrasing. But backend sophistication is not the same as a customer-facing intelligent experience, and we shouldn’t keep conflating the two.
My skepticism didn’t come out of nowhere. It comes from a pattern I’ve seen play out in my own interactions—bots that acknowledge my issue, empathize with my frustration, promise to resolve it, and then fail to deliver. Most of these are simply ineffective. One was memorable.
I recently spent 38 minutes with a bot that acknowledged my concern, empathized with my growing frustration, and kept promising an immediate solution. It never delivered one. It got stuck in a loop of reassurance with no resolution. It kept digging deeper with every promise to help. I went from frustrated to entertained, and I stayed on the line out of curiosity—would it ever recover? It didn’t. I started over and went directly to a human. Issue resolved in three minutes.
That experience was the extreme end of the spectrum, but it felt familiar. None of my other AI-enabled interactions have been that bad, but plenty have added friction instead of removing it.
Which brings me back to the original question. When we claim we’re delivering intelligent experiences through AI, whose judgment are we applying? Are we substituting our own assessment of how clever we think the system is over the customer’s judgment of whether the experience was actually intelligent? Those are very different standards, and only one of them matters (hint: the customer).
My own bar for an intelligent interaction remains high, and so far it’s only been cleared by human-to-human conversations. When I encounter an AI agent that clears it, I’ll say so and celebrate it. Until then, we should be careful and not oversell AI.
Every time we label a mediocre interaction “intelligent,” we widen the gap between our customers’ expectations and their reality—and that gap chips away at trust in the real progress AI is making.
The honest path forward is to hold ourselves to the customer’s standard, not our own—to call an experience intelligent only when the person we built it for would call it that too.
We have a long road ahead to realize the full potential of AI in customer experience. Humility about where we actually are today is the right starting point.