Most businesses are approaching AI backwards.
They start with the technology. Someone asks "what's our AI strategy?" and suddenly there are workshops, roadmaps, and a slide about transformation.
Six months later they have a deck, a pilot nobody uses, and the same spreadsheet everyone hates.
I'd start somewhere else. Start with the work.
The work is the quoting process that takes three days. The report that requires copying from five systems. The vendor PDF someone re-types into QuickBooks by hand. The handoff that breaks every time your one indispensable person takes a vacation.
Your team already knows where this work lives. Ask them what they dread doing every week and they'll tell you in about four seconds.
Why does the order matter so much?
When you start with AI, you get a solution hunting for a problem. Vague use cases. Demos that impress in a meeting and die in production. A lot of motion, no leverage.
When you start with the work, the technology questions mostly answer themselves. What should be automated. What needs real software. Where a human stays in the loop. What to build first.
A real example. A specialty contractor was turning vendor quotes into customer estimates by hand. Someone copied numbers out of PDFs, checked pricing, applied markup, entered everything into QuickBooks, and emailed the result.
It worked. It was also slow, error-prone, and completely dependent on one person.
Nobody needed an AI strategy to fix that. They needed a system that receives the quote, extracts the details, applies the business rules, and creates the estimate. AI handles the extraction and drafting. Software handles the rest. A human approves what goes out.
Notice what happened there. AI is in the system, but it was never the point. The point was getting quotes out faster with fewer mistakes.
This is more possible now than most owners realize. The build model has changed, and workflows that used to require a six-figure vendor project can become working software in weeks. But that only pays off if you aim it at real work.
So skip the strategy question. Walk the workflow instead. Watch how the work actually happens, including the steps that live in someone's head. The system you should build is usually sitting right there inside the mess.
That's what I do at Headmark AI. I build custom software, workflow automation, and AI agents for owner-led businesses and founders. No discovery phase that lasts a quarter. We walk through the workflow, find the system inside it, and get a first useful version into your hands fast. Reply to this email with the one process your team complains about most, and I'll tell you whether it's worth turning into software.