Field Note
The Integration Was Me
76% of small businesses now use AI. 14% have actually built it into how they operate.
I launched a company in the gap between those two numbers.
Coyote & Quill went live. The back office runs on AI — planning, marketing, client delivery. Not because I bought the right tools, but because I did the harder thing: I integrated them into the workflows I already run.
The reflex is to treat AI adoption as a shopping problem. That’s not where it breaks.
MIT looked at 300 deployments last year: 95% stalled. Not because the models were bad, but because no one wove them into the actual workflow. MIT called it the learning gap.
I know, because I spent six months getting it wrong.
I’ve been building my own executive planning assistant — an agent, running on Claude, meant to plan my days. For most of those six months I was quietly furious with it. If it had been a person, I’d have fired it every other week.
Four weeks ago I saw the problem, and it wasn’t the agent. It was me. I’d built the whole thing without ever teaching it how I actually work. To me my workflow is innate — I’ve done it so long I don’t think about it. But an agent can’t act on what you never say. It looks like it remembers and understands — underneath, between turns, it doesn’t.
The fix wasn’t a smarter model. It was capturing the actual workflow of my day — how I plan a deep-work session (each block gets a named deliverable, so the agent knows what “done” means), how I prep for a meeting, how I capture meeting notes, how I schedule a sprint. That’s when it started to work.
The tool took days. The integration took six months — because the integration was me.
Building a working tool fast — “vibe coding” it — is the easy part now. That’s the door. The value is on the other side: the tool woven into how you already operate, built with the person who owns the workflow, because they know where it hurts.
Big companies will figure this out on their own clock — the people, budget, time. The 30-person shop on spreadsheets and heroics doesn’t, and no one’s coming to do it for them.
Adoption was never the hard part. Integration is.
So the honest question isn’t whether you use AI — almost everyone does. It’s whether it’s built into how you operate, or just running alongside it. Which is it, at your shop?
This is one thread in a book I’m writing on military readiness models for enterprise AI adoption.
Building something with AI right now? Start a conversation — drop the one workflow you’d most want AI actually built into, not bolted beside, and tell me where it hurts. That’s where this starts.
Sources: Goldman Sachs 10,000 Small Businesses (Babson / Binder, Feb 2026) — 76% use AI, 14% fully integrate. MIT NANDA, “The GenAI Divide: State of AI in Business 2025” — 95% of gen-AI pilots stall on workflow integration, not the model.