How to find AI opportunities: follow the repetitive work
Follow the repetitive work. A one-week method to find AI and automation opportunities in your business, rank them, and pick the first one to build.
The short answer
The best AI opportunities in a small business are hiding in plain sight: the tasks people do every day, that follow roughly the same pattern each time, that nobody enjoys and everybody accepts. You do not find them by watching demos. You find them by asking your team to write down what they did this week, and then following the repetition.
The method takes one week and one afternoon.
The week: log the repetitive work
- Ask each person to keep a light log. Every time they do something for the second or third time that week, a one-line note: what, how long, which tools. Two or three minutes a day.
- Include the invisible tasks. Copying data between systems, chasing replies, reformatting documents, answering the same question, finding the file. These are the ones people stop seeing.
- Collect on Friday. One sheet, everyone’s lines, no editing. Expect forty to a hundred lines for a small team.
The afternoon: rank and choose
Score each task on three things, high, medium or low:
| Criterion | Ask | High means |
|---|---|---|
| Hours per week | Across everyone who does it | Many hours; the return is there |
| Rule share | Could you write down how it is done? | Mostly rules, a little judgment; easy to automate |
| Cost of a mistake | If the system gets it wrong once in fifty, what happens? | Cheap and noticed quickly; safe to automate |
Sort by the three scores together. The tasks at the top are frequent, mostly rule-based, and forgiving. Those are your candidates. The ones with many hours but a high cost of mistakes go on a second list: automate the preparation, keep the decision with a person.
What the list usually contains
Across the businesses we have done this with, the same items appear near the top: sorting and routing incoming email, following up on quotes, entering invoices and receipts, confirming and reminding appointments, and assembling the weekly report. We wrote them up as five processes worth automating first because the pattern is so consistent. Your list will have its own items too, and those are the interesting ones, because they are specific to how you work.
Turn the top item into a project
Measure the current hours before you build. After a month in production, measure again. That difference is the case for the second project, and the list already has it waiting.
What this means for you
Skip the brainstorm. Run the week of logging, do the afternoon of ranking, and take the top item to a scoping conversation. You will end up with a project that the team wants, a return you can measure, and a list of the next ten. That is how an AI strategy turns into work.
Frequently asked questions
Why not start from what AI can do?
Because that produces a list of impressive things nobody needs. Starting from the work produces a list of dull things everybody needs. The dull list has a measurable return and a team that wants it to succeed. The impressive list has a demo.
How much time does the week of logging take?
A few minutes per person per day: a note of each repetitive task and roughly how long it took. No tools needed, a shared sheet is enough. The point is not precision; it is to surface the tasks people no longer notice because they have always done them.
What if everything scores low?
Then your processes are already lean or your team is very small, and the honest conclusion is that there is no urgent AI project. That is a valid result, and cheaper than building something to have built something. Revisit in six months.