Where to start with AI, if you already have a business to run

Short answer: Start with a task that repeats weekly, follows the same shape every time, and that someone on the payroll dislikes doing. Measure how long it takes before you change anything. Integrate it into the tools your team already uses, and name an owner before you call it done.

Most businesses do not stall on AI because the technology is hard. They stall because the first project is chosen badly. Someone picks the most visible, most ambitious idea in the room, and it collapses under its own weight before anything ships.

A better first project is usually smaller and duller than people expect. It is the first thing we look for in a Bulletproof audit.

Start where the work repeats

Look for a task with three properties. It happens on a schedule, weekly or more often. It follows roughly the same shape every time. And somebody on the payroll dislikes doing it.

That combination matters because it gives you three things at once: enough volume for the change to be felt, enough consistency for the work to be automated, and a person who will actually help you get it right rather than defend the old way.

Reporting is the common example. So is anything that involves copying figures between two systems that were never introduced to each other.

Pick something you can measure before you change it

If you cannot say how long a task takes today, you will not be able to say whether the new version is better. Time it first. A rough number written down beats a precise number nobody collected.

This is also the honest test of whether a project is worth doing. Some tasks feel painful but only consume an hour a month. Automating them is a hobby, not a return.

Change the workflow, not the org chart

The failure mode we see most often is a tool that requires everyone to work differently on day one. Adoption dies quietly, and six months later the licence renews on something nobody opens.

Integrating into the systems a team already uses is slower to design and far faster to land. If the output arrives in the place people already look, there is nothing to adopt.

Plan the handover from the start

A system your team cannot operate without calling someone is not finished. Decide early who owns it, write down how it works, and make sure that person has watched it run before anyone declares the project done.

That is usually the difference between a pilot that quietly lapses and one that is still running a year later.