Start with one job: adopting AI without the failed project

Most AI projects fail the same way. Someone decides the business should “use AI”, a big ambiguous initiative is born, three tools get trialled, everyone is busy, and six months later there is a Slack channel full of good intentions and nothing in production.

The fix is almost embarrassingly simple. Do not adopt AI. Hand over one job.

Why one job beats a strategy

A single, painful, repetitive job has everything a vague strategy lacks. It has a clear owner who feels the pain. It has an obvious definition of done. It touches one or two systems, not twelve. And it produces a result you can look at on day one and say yes, that is right, or no, do it like this.

That last part is the whole game. AI gets good at your work the way a new hire does: by doing it, getting corrected, and doing it again. A narrow job gives it tight, fast feedback. A sprawling mandate gives it noise.

How to pick the job

Look for work that is high volume, low joy, and rules-based at the edges but messy in the middle. The inbox that gets triaged the same way every morning. The documents you chase clients for. The quotes you build from the same price book. The follow-ups that slip because everyone is busy.

Avoid, for a first job, anything where a mistake is expensive and hard to reverse, or where the “rules” live entirely in one person’s head and change by mood. Those come later, once trust is built.

Then widen, deliberately

Once the first job is genuinely useful, and people trust it, widening is easy. The teammate already knows your tools, your voice and your approvals. Adding the second job is a small step, not a new project.

This is how sensible automation actually compounds. Not one heroic rollout, but a teammate that earns a little more responsibility each month, because it kept proving it could be trusted with the last lot.

Start narrow. Get a win you can see. Then go from there.