By: Aaron Daniel, with assistance from Claude Opus 4.7
A few weeks ago I was on-site at a client, going team by team asking what part of the week each person hated most. One of their finance people described checking VAT rates across EU countries: twenty or thirty minutes a day opening a government site and copying numbers into a spreadsheet.
It's not an ROI case. No version of that justifies a proposal, a scoped build, and an integration.
So I sat down next to her, opened Claude, and said let's just build it. Half an hour later she had a single-page HTML file that pulled the data on demand. The room, including the CEO, reacted like they'd watched a magic trick.
After I'd flown home, a different employee from the same company emailed me. She'd seen what we'd done with her colleague, had her own workflow problem. We got on a call, worked through it with Claude, and she walked away with her own tool. She didn't need me to build anything. She needed thirty minutes to see how the thinking worked.
That second email reframed what I'd actually been selling. Not a tool. Not a model for building tools. The thing she'd taken from watching her colleague's screen was the start of a reflex, the moment when reaching for AI to solve a problem becomes automatic. The shorthand for that reflex is fluency. And fluency, by the time it shows up at work, was almost always built somewhere else.
The reflex
The reason corporate AI adoption is hard isn't that AI tools are complicated. They aren't. Using AI well isn't a tool skill, it's a thinking skill, and thinking skills don't get built in a workshop. They get built by use, on things that matter to the person using them.
Work alone isn't enough use. The reflex develops when you also reach for AI in the parts of your life where the stakes are low and the curiosity is yours. Last night I opened the fridge at six with no plan for dinner. Instead of standing there staring, I took a picture, sent it to Claude, and asked what I could make. Thirty seconds later I had three actual options scaled to what was in there. The point isn't the recipes, a search engine could have given me those, if I'd known what to ask for. The point is that I didn't have to know what to ask. I showed AI what I had and it figured out the rest. That's the shift. Once you've had a few of those moments, you stop treating AI like a search engine and start treating it like a thinking partner.
Once that reflex is there, it shows up at work without anyone telling you. The finance person didn't need a training program. She needed to already be someone who reaches for AI in her own life and recognizes a work annoyance as the same shape as the personal ones she's already solved.
This is why most corporate AI literacy training underdelivers. The variable isn't the workshop. It's whether the people in the room have built the reflex anywhere else. Training plus personal use compounds. Training alone doesn't.
Where this falls apart
The worst version of this idea is consulting copy that pretends it works everywhere. It doesn't.
The most common failure mode isn't the employee who's never touched AI. It's the one who has, but only for one thing. The marketer who uses ChatGPT for copywriting. The analyst who has it summarize meeting notes. From the outside they look like they've adopted AI. They haven't. They've found a single task and stopped. There's no reflex underneath, just a habit narrowed to one shape, and these people are often harder to enable than someone who's never tried, because they already think they know.
The motion is high-touch by nature, which is why it rarely gets sold. The honest pitch isn't "enable everyone." It's figure out who can be enabled, give them a real path, and accept that for some people, it isn't going to take.
How this changed how I work
The talks I give now are built around this. The architecture and ROI material is scaffolding for one live moment where I take a thing someone in the room actually does and build the tool that does it. Everything else exists to get me to that moment, because the moment is what plants the reflex.
AI isn't taking your job. The colleague sitting next to you who's actually using it might, and the path to being that colleague runs through your whole life, not just the eight hours of it you spend at work. You don't get fluent in a language by speaking it one hour a week in a classroom. AI is the same.
The question for any company starting an AI initiative isn't "what should we build first." It's "who in this room is going to be thinking with these tools by next quarter, in their work and everywhere else."