Start with AI on your dullest work, not your best idea

An AI that talks to customers on its own sounds like the big leap forward. It is also the work with the most exceptions, the highest risks and the least room for error. One wrong answer and you have a complaint, or worse. So it gets tested, tweaked and postponed endlessly, and three months later there is an impressive demo and nothing in daily use.

Meanwhile the same manual work carries on every week. Retyping invoices, sorting enquiries, writing up minutes, answering the same five questions. Nobody thinks of that work when they think of AI, because it is not exciting. That is precisely why it is there for the taking.

Dull work has three qualities that suit AI. It happens often, so every minute saved adds up many times over. It follows a fixed pattern, so you can describe clearly what good looks like. And a mistake is small and quickly spotted, because a person looks at it before it moves on. That is the ideal setting to learn how AI behaves inside your business.

A good test: find the work that makes your team sigh when it lands on the list again. Write down how often it happens and how long it takes. Start with the work that has the highest number, not the work with the best story.

Once the dull work is automated, you have something the others lack. A team used to reviewing AI output. Data already sitting neatly in one place. And time freed up to do the ambitious project properly.

The order makes the difference. Companies that start with the best idea often end up with a pilot nobody uses. Companies that start with the dullest work have something running every day within a month, and a foundation for whatever comes next.