Find Your First 3 Tasks to Hand to AI: The Hate List and the Repeatability Test
Most advice on what to hand to AI first is a list someone else made: lead scoring, onboarding, content, in their order. Your week isn't their week. This workflow starts from the tasks you already dread, runs each one through five quick questions, and leaves you with three tasks that are worth handing off, in order, plus a short list of what stays with you no matter what. It takes about an hour, and you finish with a sheet, not a vague intention.
What should a coach hand to AI first?
A task you already dread that repeats at least every two weeks, follows the same steps, takes 15 minutes or more, and that you can check faster than you can do it.
- Write a hate list. Ten minutes, every recurring task that drains you, jogged by your sent mail and calendar.
- Run the repeatability test. Five yes/no questions per task. Four or five yeses makes it a candidate.
- Cross off the hard noes. The coaching itself, a client's hard moment, your prices, and anything with confidential details you can't strip out.
- Sort by minutes per month and pick 3. No more than one of them client-facing. Hand off the first one only, for two weeks, before you touch the second.
The sheet, formulas, and an AI prompt that does the sorting with you are below.
Time check: about 45 to 60 minutes once (10 for the hate list, 15 for the test, 10 for the sheet, 15 for the handoff card on task #1). Then a 2-week trial of that one task, timed. What it saves depends on what you pick: the sheet's minutes-per-month column gives you your own estimate before you build anything. As a rough estimate, a task that passes the test usually gives back 30 to 90 minutes a week once it's running, after you count the time you spend checking AI's drafts.
Who it's for
Coaches, practitioners, and solopreneurs who've tried AI for a few one-off things and want to know where it would actually help in their own week. It's also the right first page if you're about to buy an AI tool or course and want to know what you'd use it for. If you don't know where your time goes at all, run the weekly admin time audit first and come back with that log.
What you need
- 20 minutes of your sent mail and calendar: The last two weeks of Gmail Sent and your calendar. They're how you remember the tasks you've stopped noticing. Your phone's notes app and the browser tabs you keep open help too.
- A Google Sheet: Thirteen columns, two of them formulas. Step 3 has the exact headers and formulas. Excel works the same way (SORT and FILTER need Excel 365 or Excel 2021 or later).
- Two past examples per candidate task: For each task that might pass, two real "before and after" pairs: what you started with (the notes, the email, the form) and what you finished with. You don't paste them anywhere yet. You just need to know they exist.
- Any AI chat tool (optional): ChatGPT, Claude, Gemini, or Grok, for the sorting prompt in step 4. Your hate list doesn't need client names, so leave them out. In ChatGPT: Settings > Data controls > turn off "Improve the model for everyone". In Claude: Settings > Privacy > turn off "Help improve Claude".
Example business (fictional): Tidewater Wellness Coaching
Tidewater is a made-up practice we use to show the workflow end to end. One health coach, 14 active 1:1 clients, a monthly newsletter, discovery calls booked through Calendly, invoices through Stripe. Everything here is invented to show the steps, including the tasks, the minutes, and the results.
It pairs well with Weekly Admin Time Audit: Find Where Your Hours Actually Go and AI Tool Stack Audit: Keep One Tool Per Job. New here? Get the free guide: 3 complete workflows for weekly posts, client recaps, and call prep. Not sure where to start? See every workflow by job or count your admin hours.
The steps
Write the hate list in 10 minutes
Set a 10-minute timer. Write every recurring task that you put off, sigh at, or do late at night. Don't judge them yet and don't think about AI. Write each one as verb + thing + when, so it's a task and not a mood: "write the session recap after every call", not "admin".
When you run dry, open these and keep going:
- Sent mail, last 14 days. Any email you've written more than twice with only the name changed is a task.
- Your calendar, last 14 days. What did you do in the 30 minutes after each call? That's usually where the dread lives.
- Your phone notes and open tabs. The reminder you keep snoozing and the tab you've had open for a week.
- The jog list. Copy it and tick anything that's yours:
JOG LIST (tick what you do at least monthly) [ ] recap or notes after a session [ ] prep before a session or a discovery call [ ] follow-up after a discovery call [ ] the same answer to the same question (price, availability, how it works) [ ] welcome or onboarding emails for a new client [ ] invoice and payment reminders [ ] rescheduling and no-show messages [ ] newsletter or weekly email [ ] social posts from things you've already said [ ] copying details from a form into a sheet, calendar, or email [ ] asking for testimonials [ ] sorting receipts for your bookkeeper [ ] checking which clients you haven't heard from [ ] updating your bio, sales page, or offer detailsNext to each task, give it a dread score: 1 (mildly annoying), 2 (I put it off), 3 (I'd pay someone today). If you have fewer than 8, you probably stopped before you opened your sent mail.
Done when: you have at least 8 tasks, each written as verb + thing + when, each with a dread score.
Run the repeatability test and cross off the hard noes
Dread tells you what you'd love to hand off. It doesn't tell you what AI can actually do well. For each task, answer five questions with a plain yes or no:
THE REPEATABILITY TEST (yes / no for each task) Q1 Often enough: does it happen at least every 2 weeks? Q2 Same steps: could you write how you do it in 5 lines or fewer, and would those lines be the same next time? Q3 Big enough: does it take 15 minutes or more each time? Q4 Examples exist: do you have 2 past examples (what you started with + what you finished with) you could show someone? Q5 Quick to check: could you check a first draft in a third of the time it takes you to do it yourself? 4 or 5 yes = candidate 3 yes = fix first (usually Q2 or Q4: write the steps, save two examples, then test again) 0 to 2 yes = not yetQ4 and Q5 are the two most lists skip. Without examples, AI guesses at your standard and you rewrite everything. If checking the draft takes as long as doing it, you've added a step, not removed one.
Then cross off the hard noes, whatever they scored. These stay with you in a coaching practice:
- The coaching itself: the session, the hard question, the reframe, the decision about what a client needs next.
- A client's hard moment: a reply to a client who's upset, grieving, in crisis, or unhappy with you.
- Your prices and exceptions: AI can draft the email, but you decide the number, the discount, and the refund.
- Confidential details you can't strip out: if the task only works with a client's private details pasted in, and you don't have their consent and the right settings, it waits. See AI session notes without breaking client confidentiality.
- The final send: anything that reaches a client goes out from you, after you've read it.
A crossed-off task can still have a small piece AI helps with (a checklist, a first outline). It just isn't one of your first three.
Done when: every task has five yes/no answers, a verdict (candidate, fix first, not yet), and the hard noes are marked.
Sort it in a sheet and pick your 3
Make a Google Sheet called Hand to AI. Paste these headers into row 1:
A Task | B Dread (1-3) | C Minutes each time | D Times per month | E Q1 Often (formula) | F Q2 Same steps | G Q3 15+ min (formula) | H Q4 Examples | I Q5 Quick to check | J Hard no | K Minutes per month | L Yeses | M Verdict- Fill A to D and F, H, I, J by hand, with yes or no in the question columns. Minutes are your honest guess. Round up: we all underestimate the tasks we dread.
- Formulas in row 2, then drag down:
E2:=IF(D2>=2,"yes","no")
G2:=IF(C2>=15,"yes","no")
K2:=C2*D2
L2:=COUNTIF(E2:I2,"yes")
M2:=IF(J2="yes","Keep it",IF(L2>=4,"Candidate",IF(L2=3,"Fix first","Not yet"))) - The sorted list: add a second tab called Top. In A1, paste
=SORT(FILTER('Hand to AI'!A2:M100,'Hand to AI'!M2:M100="Candidate"),11,FALSE,2,FALSE). It lists every candidate, most minutes per month first, with dread as the tiebreak. - Pick 3 from the top of that list, with one rule: no more than one of the three reaches a client. Behind-the-scenes tasks (prep, sorting, first drafts only you see) are safer to learn on. Number them 1, 2, 3.
Many first picks already have a full workflow in the library: session recaps (call transcript to client recap), call prep (discovery call prep brief), email (inbox triage agent), and payment chasing (client payment reminders). If yours is one of them, you can skip straight to that page for task #1.
Done when: the Top tab shows your candidates in order, and three of them are numbered, with no more than one client-facing.
Example: Tidewater's hate list and test (fictional)
The coach's 10 minutes, plus 5 more with sent mail open, gave 15 tasks. Here are six of them through the test (minutes are her own guesses):
Task | Dread | Min each | Per month | Q1 Q2 Q3 Q4 Q5 | Hard no | Verdict
Write session recap after each call | 3 | 20 | 40 | Y Y Y Y Y | no | Candidate (800 min/month)
Prep before each discovery call | 2 | 15 | 8 | Y Y Y Y Y | no | Candidate (120 min/month)
Monthly newsletter from my notes | 3 | 120 | 1 | N Y Y N N | no | Not yet (2 yes)
Answer "how much is coaching?" DMs | 2 | 10 | 12 | Y Y N Y Y | no | Candidate (120 min/month)
Reply to a client who missed 3 sessions | 3 | 25 | 2 | Y N Y N N | yes | Keep it
Copy intake form answers into my client sheet | 2 | 15 | 4 | Y Y Y Y Y | no | Candidate (60 min/month)Her pick: 1. session recaps (behind the scenes until she sends, and the biggest by far), 2. discovery call prep, 3. copying intake form answers into her sheet. The pricing DMs tied with call prep at 120 minutes, but they reach a prospect directly, and recaps already were her one client-facing pick, so they waited. The missed-sessions reply was a hard no: that's a client in a hard moment, and it stays hers.
Where this breaks
- The list is all feelings, no tasks. "Email", "marketing", "admin" can't pass a test. Fix: verb + thing + when, and split anything that's really three tasks.
- You guess the minutes low. Dreaded tasks feel shorter in memory because you do them fast and late. Fix: round up, and time it once by hand before the trial if the number decides your pick.
- You pick the task you'd love to be rid of, not one that passes. The newsletter you dread most may fail Q5 because only you can tell if it's good. Fix: dread is the tiebreak, not the first sort.
- You hand off three at once. Three new Projects, none tuned, all mediocre. Fix: task #2 starts only after #1 is a keep or a drop.
- No examples, so AI guesses your standard. Generic drafts, a full rewrite, and the conclusion that "AI doesn't work for me". Fix: Q4. Save two real before-and-after pairs first.
- Checking time isn't counted. It feels faster because the typing is gone, but 15 minutes of fixing a 20-minute task is 5 minutes back. Fix: the trial log has a checking column for a reason.
- A hard no slips in. A client reply in a hard moment gets drafted by AI because it scored 4 yeses. Fix: the hard-no column wins over the score, always.
Start with 3 complete workflows, free
Steps 4 to 7, including the second-opinion AI sort prompt, the one-page handoff card template, the 2-week trial log with keep, fix, and drop rules, the monthly hand-off check prompt, and Tidewater's full example, live in the members library. To try the approach first, grab the free guide: 3 complete workflows for weekly posts, client recaps, and call prep.
Check your inbox and tap the confirm link to get your guide. Free, plus one new workflow idea a week. Unsubscribe anytime.
Want this full workflow now? Founding members get the whole library for $19/mo, locked in. First 25 members, closes Oct 29.
*Time saved is an estimate based on typical manual effort for this task, not a measured result. Your mileage will vary. Always review AI output before sending or publishing.