CaretWhat to delegate as a startup founder first, ranked
It's 11:40 on a Sunday night and you're doing the thing you swore you'd stop doing. You're copying last week's signups from one dashboard into a sheet so you can paste a screenshot into the team channel on Monday. It takes forty minutes. Nobody will read it closely, and if you skip it, nobody else will do it.
Most delegation advice assumes you have someone to delegate to. At two to fifteen people you usually don't. Everyone already does two jobs, and the work that falls to you is the work without an owner.
So the useful question is narrower: which work should leave your hands first, and which work should never leave them?
The test: three questions for every chore
Run each chore through three questions before you rank anything. For a whole week of chores it takes ten minutes.
- Does it repeat? Weekly or monthly work pays back the setup. A one-off doesn't.
- Can you say what "done" looks like in one sentence? "A sheet with last week's signups, revenue and churn, posted Monday 9:00" is a spec. "Make our marketing better" is a wish.
- Can you check it in under five minutes? If reviewing the result takes as long as doing it, you haven't saved anything yet.
Three yeses means hand it off now. Two means hand it off with a template. One or zero means it stays with you for now.
The ranked list
This is the order I'd hand work off in, at a startup with no ops, finance or marketing hire.
| Rank | Work | Why it goes early | What you still do |
|---|---|---|---|
| 1 | The weekly numbers report | Same shape every week, easy to check against the source | Read it and decide what to do about it |
| 2 | Chasing failed payments | Money on the table, a clear list, polite templates work | Approve each email before it goes |
| 3 | Research before calls and meetings | Reading a company's site and news is slow and low-judgment | Pick the questions you want answered |
| 4 | Follow-ups after calls | The notes exist; the draft is mechanical | Fix the tone and hit send |
| 5 | Tracking who promised what | A list of owners and dates, rebuilt from the week's threads | Chase the people, which a list can't do |
| 6 | First drafts of posts, updates and docs | Blank-page time is the expensive part | Rewrite the parts that need your voice |
| 7 | Site and SEO chores | A checklist with right answers (titles, meta, broken links) | Approve what ships |
| 8 | The investor update draft | Mostly assembly from numbers and threads you already have | The story, the asks, the honest bits |
Look at the right-hand column. You don't disappear from any of it. You move from doing the work to checking it, and checking is where your judgment counts.
The five things to keep
Some work fails the test on purpose. These stay with you until you hire someone you'd trust with the company:
- Hiring and firing. Delegate the sourcing sheet, never the decision.
- Pricing. A draft pricing page is fine. The number is yours.
- The hard customer conversation: the churn call, the outage apology, the "we can't build that".
- Fundraising relationships. An assistant can prep the update. Investors want to hear from you.
- The final yes on anything external. Emails, posts, payments, deploys. Someone gets to be wrong in public, and at this size it's you.
An example: handing off failed payments in one afternoon
Rank 2 is a good first hand-off. The numbers here are an example, not a real customer.
The ask. "Find every failed payment from the last 30 days and draft a short, friendly email for each one. Don't send anything."
What came back. A sheet with nine rows: customer, plan, amount, failure reason, date, and whether the card had been updated since. Total at risk: $3,670. Under it, nine draft emails, each with the customer's name, the amount, and a link to update the card.
What the founder changed.
- Removed two rows. Both customers had already said on a call that they were cancelling, and a dunning email would land badly.
- Rewrote the first line of the template. The draft opened with "We noticed your payment failed", which reads like a robot. The new line: "Your card didn't go through for October. Probably an expired card, happens to everyone."
- Added one rule for next time: never email a customer twice in five days.
What got approved. Seven emails, sent after one review. Most of the founder's time went on the two customers who were leaving. The sheet stays, so next month starts from it.
That's the pattern for everything on the list: the hand-off saves the assembly, and your time goes to the judgment calls.
find every failed payment from the last 30 days and draft an email for each. Don't send anything.
Attached failed-payments.xlsx: 9 rows, $3,670 at risk. Nine draft emails below, each with the amount and a link to update the card.
drop the two who said they're cancelling. Open with "Your card didn't go through for October." And never email anyone twice in five days.
Done. 7 emails ready, rule saved for next time.
Send 7 emails
ApproveDenyApproved.
An example thread, shortened.
How we do it at our company
Five bots do most of our day-to-day work: a chief of staff, an engineer, a designer, a marketer and a generalist. Each one took a column from the list above. The chief of staff keeps the to-do list and the weekly summary. The marketer did our blog, launch copy and SEO chores. The engineer reviews and ships code. A person, usually me, says yes before anything goes out.
How to set it up in Caret
- Sign up at app.caretbot.com. Team starts with a 7-day trial.
- Pick the one chore from the ranked list you did three times last month. Create a bot for it and write its job in plain words, including what it should never do on its own.
- Connect only the apps that job needs, through connected apps. A payments bot doesn't need your calendar.
- Add the bot to a team thread. Threads are bridged to Slack both ways, so it can live where your team already talks.
- Give it a schedule for the repeating part, for example "Monday 09:00".
- Approve or deny when it asks. Before a bot sends, posts, pays, deploys or deletes, it posts an approval card, and an admin answers with one click.
- When a run goes well, ask it to save the steps as a skill so next month starts from the same procedure.
A few asks to start with:
- "every monday 09:00 put signups, revenue and churn for last week in a sheet and post it here"
- "what did we say we'd ship this week, who owns each thing, and what's still open"
- "brief me on the company I'm meeting at 3: what they sell, who's on the call, anything in the news"
Caret's Team plan is $40 a bot a month, with a 7-day trial. Each bot works on its own computer, so bots keep working when your laptop is closed, and the whole team is in the threads.
FAQ
What should a startup founder delegate first? Recurring work with a clear finished state that you can check quickly: the weekly metrics report, failed-payment follow-ups, and research before calls. These pay back the setup in the first week and carry little risk because you review the output.
What should a founder never delegate? Hiring decisions, pricing, the hardest customer conversations, investor relationships, and the final approval on anything that goes out under the company's name. You can delegate the prep for all of them.
Can I delegate to AI before I can afford to hire? Yes, for work that passes the three-question test. Keep the approval step on anything external, and give the AI a written job with the things it must not do alone.
If you want to see how this looks for a whole founder week, read what Caret does for founders. Put a bot on it: start your 7-day trial or download the Mac app.