Every business I talk to has been told it needs an AI strategy. Most of them have bought a few licences, run a lunch-and-learn, and watched nothing change.
That isn't because AI doesn't work. It's because "use more AI" isn't a plan. It's a tool looking for a job.
Start with the job, not the tool
The useful question isn't "where can we use AI?" It's "where does the time go?"
Walk round the business and look for work that is:
- Repetitive. The same kind of thing, many times a week.
- Rule-shaped. A new starter could do it from a written checklist.
- Text-heavy. Reading emails, documents or forms and pulling out what matters.
- Low on judgement. Getting it slightly wrong is annoying, not a disaster, and a person can check it.
Jobs like that are where AI earns its keep: sorting an inbox, reading delivery notes, drafting the first version of a quote, answering the questions your team asks each other every week.
Where it doesn't belong
Some work should stay with people, and pretending otherwise is how AI projects get a bad name:
- Anything that needs a relationship. The difficult customer, the supplier you need a favour from.
- Anything you'd be embarrassed to get wrong in public.
- Decisions about money, people or direction. AI can lay out the options. It shouldn't choose.
A small first job
Pick one job that eats hours every week and fits the list above. Write down exactly how it's done today. Then build the smallest thing that takes the boring part away, with a person checking the output.
Measure it against the before. If it saved the time, do the next one. If it didn't, you've lost a few weeks, not a year.
That's not an AI strategy. It's better than one.