The usual advice for a side hustle next to a full-time job is to “learn AI” and pick a niche later. That is backwards.

If you only have four honest hours a week, a general “AI consultant” offer will not fit. Neither will a custom app, an unbounded retainer, or a service that requires you to answer clients at 2pm while you are employed.

What does fit is one repeating problem a small business already feels — missed enquiries, unsent quotes, Sunday-night spreadsheets, reviews that never get asked for — turned into a small, reviewed artefact they can inspect.

The filter I use:

• Can I name the owner and the weekly leak in their language, without saying “AI”?

• Can I deliver the result inside protected evening or weekend blocks, including time to check the work?

• Can I reach ten real people in that type of business without running ads?

• Would I still do the first engagement if nobody clapped?

If any answer is no, the idea is a hobby. Rejecting it is the work.

I built Side Hustle Set-Up around that sequence: map your actual week, find a problem with commercial urgency, commit to one minimum viable service, package it so a buyer understands it in a minute, then test it with a controlled first client.

Tools come last. Proof comes from a real week, not from a new model launch.