The 90-minute AI audit: find the task costing your team time

A bigger AI plan is usually not the starting point

Most small businesses do not need a twelve-month AI roadmap. They need a clearer view of the work that keeps pulling someone away from customers, delivery, or sales.

That is what a useful AI audit is for. It is a working session that names one repeated problem, measures its current cost, and decides whether a small system could make a meaningful difference. It is not a tool demo. It is not a maturity assessment that ends in a slide deck.

The aim is simple: leave with one prioritised opportunity and a sensible next step.

Start with the interruption, not the software

Ask three questions.

1. What did we do repeatedly this week that did not need much judgement?
2. Where did a customer have to wait because everyone was busy?
3. What work exists only because two systems, or two people, do not connect cleanly?

The answer might be replying to the same enquiry, copying details into a spreadsheet, chasing a lead, creating a weekly update, or finding a previous answer in a shared inbox. The best first opportunity is rarely the cleverest one. It is the one people already recognise as annoying and expensive.

Put a number beside the problem

You do not need perfect data. A useful estimate is enough. Note how often the task happens, how long it takes, who does it, and what happens when it is delayed or missed.

For example, a business might spend four hours each week responding to routine booking questions. The work could be partly reduced with a clearer website flow, automatic confirmation, and a human hand-off for exceptions. The value is not that an AI tool can write an answer. The value is that the owner stops being the switchboard.

Keep time savings and possible revenue separate. A missed enquiry has a different value from an hour of internal administration. Treating them as one impressive number makes the decision less trustworthy.

Choose the smallest useful test

A good audit ends with a test, not a promise. Define the job, the input, the output, the human owner, and the point where a person takes over. Then run it for two weeks on real work.

If the new step saves time and still feels safe, write it into the existing workflow. If it does not, remove it. That is progress too. The goal is not to make AI visible in the business. It is to make the work calmer.

What to bring to an audit

Bring one messy process, a recent real example, and the person closest to the work. Leave the shopping list of AI tools at home. Once the job is clear, the right technology is much easier to judge.

If you are not sure where to begin, start with the interruption your team complains about most often. It is usually trying to tell you something.

Next step: Book a 20-minute AI opportunity audit and bring the bit of work that keeps interrupting the real work.