How to implement AI in a company

To implement AI in a company, start with a specific, repetitive and measurable task, try it for a few weeks with a tool that already exists and only then decide whether it deserves more investment. Doing it the other way round, buying a platform or hiring a big project from the outset, tends to be expensive and leaves half the team not knowing what it's for.

A few simple steps: make a list of the tasks that repeat, pick one and one person to try it, set what data can't go into the AI (customers' personal data, contracts, passwords), measure how long that task takes before and after, and review the results. What usually confuses people is thinking implementing means developing something: at first, most small businesses don't need to program anything. And don't forget the people: if the team doesn't understand what it's for, they won't use it. With personal data, keep data protection in mind; when in doubt, better to ask someone who knows.

This is the way of working that convinces me most in small businesses: little by little and measuring. In "Artificial intelligence in your business without launching a project" I describe the three uses I see working in real small businesses, how to get them running in a week and what return they give, before spending on consultancy or development.