Keep this open while you do the work. It covers three moves: pick a first task, log the proof in numbers, then propose an AI role built on what you can show.
The two-box filter
Write out what you actually do in a week: the recurring tasks that come back every Monday, not the big projects. A task is fit to start with only if it passes both boxes at once:
- [ ] It eats real hours every week. A task that costs you fifteen minutes a month gives you nothing to point to later.
- [ ] A slip is cheap. You stay in the loop, check the output, and fix it before anyone downstream sees it. If one mistake reaches a customer or a number in the books, set that task aside for now.
One box without the other isn't enough. Tasks that usually pass both:
- [ ] Your weekly status report
- [ ] Meeting notes
- [ ] Sorting your inbox
- [ ] Cleaning up data in a spreadsheet
- [ ] Basic research behind a single decision
Proof log
Automate the tasks at the top of your list, then write down the numbers before you forget them. Count the whole task, including the check on whatever the AI hands back, not just the second you click "generate." Fill one row per task.
| Task | Time before | Time after | Hours/week saved | What the AI does now |
|---|---|---|---|---|
| Weekly report | 2 hrs | 10 min | ~1.8 hrs | Pulls the data and drafts it |
Visibility checklist
The log won't convince anyone while it sits in your folder. Make the proof visible.
- [ ] Demo one win at a team meeting, briefly, on a single example.
- [ ] Offer to fix a coworker's most annoying task, the one they were complaining about.
- [ ] Document each workflow as you go, so someone without your practice can repeat it.
- [ ] Collect your best prompts and workflows in one shared internal doc the team can use.
Say it in business terms. How you talk about a win weighs as much as the win itself:
Instead of: "I used AI for this." Say: "This saved us 8 hours before the quarterly report."
From annoyances to bottlenecks
Once you have a few wins and people come to you for help, run the audit again, this time across the whole business. The question changes from "what annoys me" to what actually holds the company back.
- [ ] If we doubled our customers tomorrow, what would break first?
- [ ] Where does work pile up and wait, no matter who is on it?
- [ ] Which step does every project snag on before it can move?
- [ ] What do we lose customers or hours to that nobody officially owns?
Where the answer lands is your first real project.
One number
Roll all the hours and money your automations have saved into a single number. Scattered favors are hard to budget; one number is a line a manager can plan around.
Across my automations, I'm giving the team back the equivalent of one full-time hire a year.
Then take that number to your manager and propose a concrete role, with a title, built on what your log shows. You don't ask for the role. You show you're already doing it.
Regulated industries
Don't point AI at sensitive data and don't automate anything without sign-off. Practise on the side with stand-in data instead.