There's a scene I keep watching in different companies. The founder shows me the new AI assistant with the pride of a parent at a school concert. Look, it understands context. Look, it answers at two in the morning. Look how innovative we are. And I believe them — the technology is genuinely impressive. Then I ask the only question that matters: did anything get better for your customer? And the room goes a little quiet.
Because here is the uncomfortable truth every business eventually meets: the customer does not care that it is AI. Not positively, not negatively. The customer arrived with a problem — a double charge, a missing order, a question about whether this thing fits their situation — and they will judge whatever meets them at the door by one standard only: did it solve my problem without wasting my time? Novelty is your emotion, not theirs. Nobody has ever thanked a company for the sophistication of the thing that failed to help them.
I come at this from a slightly unusual angle. I spent years evaluating government services — the least glamorous service sector on earth — and if that work teaches you anything, it's that customers everywhere apply the same brutal arithmetic. When a government office digitized a permit process, citizens didn't ask what technology stack powered the portal. They asked three numbers' worth of questions: is it faster than before, is it fewer steps than before, does it cost me less — in money, in paperwork, in nerves? If yes, adoption followed without a single poster. If no, no amount of "modernization" messaging moved anyone. I've carried that three-number test into every AI conversation since, and it embarrasses expensive projects with remarkable consistency.
Run it on your own assistant, honestly. Take a real customer request from last month — not the demo scenario, a real one, preferably one that ended badly. Walk it through your bot with a stopwatch. Faster than your old process, wall clock to resolution — not to first response, to resolution? Fewer steps for the customer, counting every repetition, every "let me transfer you," every form? Cheaper for the customer in effort and nerves — including the two hours they spent fuming while your dashboard showed a three-second response time? Your instant is not their instant. The customer's clock starts when the problem appears and stops when it's gone; everything between is cost.
If the honest answer is "it's about the same, but automated" — you haven't improved your service. You've relocated your queue into a chat window and called it transformation. The customer notices. They always notice. They just don't file their noticing under "feedback about AI"; they file it under "this company got harder to deal with," and then they file themselves under someone else's customer base.
Now, a nuance, because I've published a whole kit about disclosure and someone will reasonably ask: if customers don't care that it's AI, why insist on telling them it's AI? Because those are two different sentences. Disclosure is not marketing — it's the entry ticket. In the EU it's now the law; in a growing list of US states, likewise; and everywhere it's simple respect — people are entitled to know whether they're talking to a person before they pour their situation out to a machine. But the ticket only gets you into the theater. It doesn't earn applause. "Powered by AI" convinces no one of anything, in the same way "under new management" never made anyone's lunch taste better. Tell them honestly, then win them on the only field that counts: the outcome.
The good news hiding in all this: indifference cuts both ways, and the second way is an enormous gift. Customers who don't reward AI for being AI also don't punish it for being AI — the moment it's genuinely better, they switch loyalties without a flicker of sentiment. People who swear they "hate chatbots" cheerfully use the one that reverses a charge in ninety seconds at midnight. What they hated was never the bot. It was the wasted time wearing a bot costume. Beat the human process on the three numbers and adoption takes care of itself — no education campaign, no "meet our AI" newsletter, no innovation theater required.
So my advice runs opposite to the instinct of almost every launch announcement: stop selling the AI. Sell the fifty minutes the customer gets back. Measure yourself by their stopwatch, not your model card. In the end, the technology in the middle is your business — literally your cost line, your vendor choice, your risk to manage. The outcome at the end is theirs. They will judge you at that end, and on nothing else. Meet them there.