Which first job to give AI in a small firm
The best first job for AI in a small firm is a small, boring task the team repeats every week, where a person can check the answer in a few minutes and a mistake costs nothing. Sorting customer reviews, tidying a lead list and summarising the weekly report all fit.
Replying to a customer, approving a bill or changing a price don't fit yet, because a mistake there reaches someone outside the firm. This post gives six first jobs that work well, the tests behind them, and what to set up before the AI starts.
What makes a job safe to start with
Four tests decide it. The job repeats at least weekly, so the team gets enough tries to learn from. A person can check the output against something real, such as the original reviews or last month's report. Nothing leaves the firm without a person reading it first, and the data in the job isn't on the firm's never-paste list, which a separate post on this site covers.
The last test rules out more than owners expect. A salary summary is a repeated, boring job, but it carries every employee's pay, so it waits until the firm has a business plan and a clear rule for it.
Six first jobs that pass the tests
- Sorting customer reviews. The AI reads last month's reviews, with names removed, and groups them by what people praised or complained about. A person checks ten reviews against the groups to see whether the sorting holds.
- Tidying a lead list. The AI cleans the company columns of a list: one spelling per company, one column for city, duplicates flagged. A person checks every flagged duplicate before anything is removed, and phone numbers stay out of the chat.
- Summarising the weekly report. The AI turns the weekly sales or operations report into five lines for the owner. The person who wrote the report checks the five lines against it, because that person knows which number matters this week.
- First drafts of routine documents. A job description, a supplier enquiry or a meeting agenda gets a first draft from the AI. A person edits it and sends it, so the AI saves the blank-page time and nothing else.
- Meeting notes into an action list. From notes the team already typed, the AI lists who does what by which date. The person who ran the meeting checks the list before sharing it.
- The second version nobody had time to write. Many firms have an offer email or a product page they meant to test against a second version. The AI drafts that version in minutes with a note on what it changed, and a person picks which one the firm tries.
Three things to set up before the AI starts
- A job card. It names the job, the hours it costs the team each week, and what the AI must never do in it. The hours matter because they're how the firm will know later whether the job was worth handing over.
- A watcher. One named person checks the output every time for the first month. The AI never checks its own work, because it marks a wrong answer correct in the same confident voice it uses for a right one.
- A one-page report at the end of the month. It says what the team expected, what happened, and the rule the firm is keeping. Most firms skip this page, and it's the page that makes the second job faster than the first.
What I got wrong in my own work
I run my own firm's daily work with AI helpers, and I made the mistake this post warns about. For most of a year I counted how much the helpers did, and that number went up every week while I stayed exactly as busy as before. Then I changed what I counted to how often a helper needs me to step in. These days it's about four times a week, and each step-in either becomes a written rule or stays a call that only I should make.
Where this advice stops
This is a way to pick a firm's first few jobs, and it doesn't cover everything AI can do for a firm. Once three or four small jobs run well with a watcher, the firm can try bigger ones, such as a helper that holds any supplier bill above a limit the owner sets. That step needs more setup, and it shouldn't be the first one.
Picking the first job is easier with the whole team in the room, because the people who do the work know which tasks they repeat. That's what the hands-on workshop I run with small leadership teams covers. You can ask about the workshop here.