Where to actually start with AI in your business

Everyone wants to "do something with AI." Far fewer people can tell you which specific job they want it to do. That gap is where most of the money and enthusiasm gets wasted.

So when a business asks me where to start, I don't start with the technology at all. I start with a simpler question: where does your team lose time to work that nobody enjoys and nobody would miss?

Look for the boring stuff first

The best first AI project is almost always unglamorous. Summarising enquiries. Drafting first-pass replies. Pulling figures out of documents. Tagging and routing things that arrive in an inbox. None of it makes a good headline, but all of it quietly eats hours every week.

Start there, because:

• The task is well understood, so you can tell immediately whether the AI is doing it well.
• The downside of a mistake is small and easy to catch.
• The person who does that task today can judge the output far better than any consultant.

Keep a human in the loop

Early on, AI should draft and a person should approve. That one rule removes most of the risk and almost all of the fear. Your team stops worrying about being replaced and starts treating the tool like a fast, tireless junior assistant — which is exactly what it is.

Measure the hours, not the magic

A good first project has a number attached: this used to take four hours a week, now it takes forty minutes. If you can't express the win in time or money, you've probably picked something too vague.

Get one of these working, let the team feel the relief, and the appetite for the next one takes care of itself. That's how AI actually enters a business — not with a grand strategy, but with one annoying job quietly disappearing.