Where to start with AI in a small business: a five-question audit
Before you buy another AI tool, answer these five questions. They'll tell you where AI will pay off in your business, and where it won't.
Most small businesses don't have an AI problem. They have a where-do-we-even-start problem. There are a thousand tools, everyone on LinkedIn says you're behind, and nobody has time to run experiments on top of the actual job.
So here's the shortcut we use in the first conversation with every client. Five questions. Answer them honestly and you'll know where AI belongs in your business, or whether it belongs there at all yet.
1. What work do you do every week that follows the same steps?
AI is at its best on work that's repetitive but not quite mechanical: reading an inbound email and deciding what it is, pulling details out of a document, drafting a first version of something you've written a hundred times. Make a list. If it happens weekly and follows a recognisable pattern, it's a candidate.
2. Where do things fall through the cracks?
Missed follow-ups, documents that never arrived, a client who waited four days for an answer. These are often more valuable to fix than the "big" tasks, because each one costs you trust as well as time. An agent that never forgets is a surprisingly powerful thing.
3. Who is the bottleneck, and what are they doing that someone else could?
In a 5–50 person business there's usually one or two people everything routes through. Look at their week. Every hour of their time you free up is worth more than an hour of anyone else's.
4. Where does your data actually live?
AI needs something to work with. If your client information is spread across inboxes, a CRM nobody updates and three spreadsheets, that's fine. It's normal. But it shapes what's realistic. Sometimes the first win isn't AI at all; it's connecting the tools you already have so there's one source of truth.
5. What would "worth it" look like in numbers?
Hours saved per week. Faster response times. More leads followed up. Pick the one metric that would make you say "that paid for itself". If you can't name one, you're not ready to spend money yet, and a good consultant should tell you that.
The best first AI project is boring, frequent and measurable.
What to do with your answers
Rank your candidates on two things: how often the work happens, and how painful it is when it goes wrong. The one in the top-right is your first project. Keep it small enough to be live in weeks, not months, and make sure a human stays in control of anything customer-facing.