A while back, I built an AI tool to write my team's weekly sales email.

It was good. I'd fed it our positioning, our recent wins, the competitive landscape, the things reps actually needed to know before a call. It produced something clean, well-organized, and genuinely useful — exactly the kind of thing I thought a busy sales team was missing. I was a little proud of it.

My own reps stopped reading it.

Not all at once, and not because it was wrong. The content was strong. But it was long, and it was one more thing to read for people who were already drowning in things to read. So I did the unglamorous thing: I cut it down. Three lines and a link. The single most important update, and a place to click if you wanted more. Open rates came back the next week.

I think about that small, slightly embarrassing failure more than almost anything else I've built. Because it taught me the thing I now organize my whole working life around:

In enterprise AI, capability was never the constraint. Attention is. Adoption is.

We have spent two years marveling at what these models can do — and they are genuinely remarkable. I've personally built two working apps for the cost of two monthly subscriptions, with no engineering team and no budget. A few years ago that sentence would have been science fiction. The raw capability now sitting on every knowledge worker's desk is staggering.

And yet, walk into most large companies and you'll find the same quiet story: a six- or seven-figure AI investment, a proud announcement, a pilot — and then very little. Usage flat. The tool technically "live" and functionally ignored. Leadership frustrated, because they did the thing everyone told them to do and the payoff never arrived.

Here's the uncomfortable truth I keep running into: the model is the easy part now. You can buy the best AI on the market and still get nothing back, because a tool nobody adopts is just a line item. The hard part — the part almost every rollout skips — is whether real people fold the thing into how they already work. That's not a technology problem. It's a human one. It's about habit, trust, attention, and fear. And it is wildly under-served, which is exactly why it's where the leverage is.

I didn't arrive at this conviction through a data science background. I arrived at it through a career that makes very little sense on paper and has, in practice, been about the exact same problem the entire time.

I started in a hospital, as a patient and family advocate in a level-one trauma center. When the helicopter landed, part of my job was to find out who the patient was, reach their family, and walk that family through the worst hour of their lives — sometimes including the news that the person they were waiting for wasn't going to make it. You learn things in a room like that you can't learn anywhere else. Mostly you learn that information is the easy part. How a person receives it — whether they can actually absorb it and act — depends almost entirely on the human delivering it.

That, improbably, is why Apple hired me. This was back when iPhones weren't waterproof and there was no automatic cloud backup. People would drop a phone in a sink and lose every photo of their kids, and they would come to the Genius Bar and cry — right there in a retail store. Apple didn't bring me in to fix computers. They brought me in because of the hospital: to teach brilliant technical experts how to deliver hard news with empathy, how to sit with someone in a small frustration that felt enormous to them. They taught me the technology; I coached the humans.

Then came seven years at ServiceNow. First as an admin and developer — I actually built things, and I still do. Then as a seller carrying a multi-million-dollar number, advising CIOs, CFOs, and transformation leaders on how to make AI real inside enormous, complicated organizations. Same lesson, new vocabulary: get the human part right, and the technology finally sticks.

Four very different rooms. One repeating pattern.

If the human side is the hard part, the natural question is: so what do you do about it? Here's the discipline I keep coming back to, because it's the one that consistently separates the AI projects that pay off from the ones that quietly die.

Start from the business result, not the tool. Not "where can we use AI," but "what specific number are we trying to move, and who is accountable for it?" If you can't name that, you're not ready to buy anything yet.

Work backward to the use case. Once you have the outcome, the use cases that matter become obvious — and so do the dozens of shiny ones that don't.

Ship a small, real version fast. Not a year-long transformation. An MVP that one team uses next month, so you learn from reality instead of slides.

Then — and this is the part everyone underestimates — change the process around the tool, not just the tool. The biggest adoption blocker I've ever seen wasn't training or technology. It was that the workflow the AI was supposed to help never actually changed, so the tool became extra work bolted onto old work. Redesign the process and usage takes care of itself.

And govern from day one, so that as it scales it stays funded, measured, and trusted.

None of that is about the model. All of it is about people and the systems they work inside. That's the whole point.

I write this for the people doing the hard version of this work.

The CIO whose board wants AI results and whose pilots keep stalling. The leader who can feel the ground shifting under their industry and wants to get ahead of it — without turning their company into a layoff story, because I genuinely believe the best version of this future is a capability play, not a headcount-reduction play. And the individual who isn't running a transformation at all, who just wants to be a little better, a little more capable, this week than last.

That last reader is the whole spirit of the name. The Increment. I don't believe in the overnight transformation. I believe in getting meaningfully better a little at a time — and in the quietly radical fact that one motivated person with these tools can now do what used to take a team.

I should tell you up front: I'm an optimist about this, openly and probably to a fault. I think this is the most empowering moment to be a curious, motivated person in a generation. The barrier to building, learning, and creating has collapsed, and most people haven't noticed yet because they're waiting for permission or paralyzed by the noise.

The technology will keep getting easier. The human side — attention, trust, habit, the courage to actually change how you work — is where the leverage is, and honestly, where the fun is.

If that's your kind of problem, I'd love to have you along. I'll share what I'm building, what I'm learning about selling and adopting AI inside real organizations, and the occasional honest failure like the sales email.

So let me ask you the question I'll keep coming back to: where have you seen good technology quietly ignored — and what do you think it would have taken for people to actually use it?

I read every reply.

— Isaac