Falling in love with the problem
Why most AI tools don’t stick—and what to do about it
There’s a lot of pull for AI right now. Everyone wants the productivity gains. The promise is real: faster execution, sharper insights, infinite interns in a box.
And yet… adoption lags.
Inertia is real—especially in orgs where decision cycles are long, tooling is fragmented, and incentives are misaligned. If you’ve ever sold to a mid-size team, you’ve probably heard some version of:
“We’re excited... but it’s not a top 3 priority.”
The truth is: productivity alone rarely drives adoption.
Not unless the experience is 10x better than the status quo.
Not unless it solves a problem someone is already trying to fix.
And that’s the core tension in AI right now: the tools are advancing, but they’re often solving the wrong thing.
Technologists fall in love with the tech. Founders fall in love with the problem.
This is a trap I’ve fallen into myself. The models are extraordinary. The potential is enormous. It’s easy to start with the tech and work backwards.
But if you want to build something sticky, you have to start with the friction. What’s hard? What’s clunky? What slows people down—or burns them out?
AI unlocks are only valuable if they slot into something people already do—or desperately wish they could do better.
You don’t just need a demo-worthy prototype.
You need deep empathy with the job to be done.
I’ve been talking to operators and CTOs who admit they still haven’t built an AI habit.
Not because the tech isn’t impressive—but because they haven’t found a real wedge into their workflow.
Sometimes, the problem is friction:
“It gave me mediocre output and I didn’t know how to improve it.”
Other times, it’s team dynamics:
“I’d use this, but I collaborate across so many people and tools—it just doesn’t fit.”
And more often than you’d think, it’s institutional blockers:
“We can’t use AI at all internally. Our company is worried about IP risk.”
One physician friend even told me he texts all his ChatGPT queries to a CTO buddy—so he doesn’t compromise his own privacy.
The takeaway? The why behind slow adoption isn’t always obvious. But it’s always real.
That’s why flashy demos don’t drive growth.
Real usage comes from real insight.
So what should you build?
AI tools that stick usually get four things right:
✅ They solve a painful, frequent problem
✅ They plug into existing tools and team habits
✅ They let the user feel more competent, not less
✅ They protect trust — around data, privacy, and control
In this new wave of tooling, it’s not enough to show someone what’s possible.
You have to make their job feel easier, safer, and more powerful.
That starts by falling in love with the problem — not just the tech.
What’s one messy problem your team has fallen in love with—and actually solved with AI?
👇