All guides

Start here · Lesson 1.5

10 Ways to Decide What to Automate With AI This Week

Rafay Korai · 6 min read · 2026-11-04

There is plenty of advice on what AI can do and very little on what to automate with AI in your own week. Generic lists of ideas do not help, because your week is not generic. What helps is a handful of ways to look at your own work and spot the jobs worth handing over.

Below are ten tactics. They are not a sequence and you do not need all of them. Read through, pick two that suit your situation, and try them this week. If you want the full guided version, Lesson 1.5 covers the method end to end: Find the jobs to hand to AI.

Look at what you already repeat, pick the task you can explain in one page, and test it on last week's real work before you rely on it.

1. Read your sent folder backwards

Open your sent email and scroll through the last two weeks. Every time you see a reply that looks almost like another one, mark it. Repeated replies are a pattern, and patterns can be described in a brief. This works well for quotes, meeting follow-ups, booking confirmations and polite refusals.

2. Look at your Friday afternoons

The jobs you push to the end of the week are often the repeating admin that never feels urgent: invoices, reports, updating a spreadsheet, tidying files. Write down what you do on Fridays for three weeks in a row. Anything that appears twice is a candidate.

3. Count the questions you answer twice

Check your messages, comments and customer emails. If you have answered the same question more than twice, you have a job that is either a saved reply, a short page on your site, or a draft AI can prepare for you to approve. Count them. The question that appears most often is usually first in line.

4. Use the one-page test

Could you explain the task to a new hire on a single page? If yes, you can brief AI on it. If you would need to say "it depends" every other line, the task needs your judgement and should stay with you for now. Anthropic's guidance suggests a similar check: show your instructions to someone with no context, and if they would be confused, the AI will be too.

5. Score before you pick

Give each candidate a score from 1 to 3 on how often it repeats, how long it takes and how easy it is to hand off, then add them up. It sounds mechanical, and that is the point. It stops you picking the task that annoys you most instead of the one that pays back most.

6. Prefer jobs with a checkable result

A good first job is one where you can look at the output and know in a minute whether it is right. A draft summary of meeting notes you attended is checkable. A forecast of next quarter's sales is not. Choose outputs you can judge quickly, because checking is part of the job.

7. Start with work only you will see

Internal drafts, notes to yourself and first versions of documents are low risk. If the AI gets something wrong, the mistake is caught before it reaches a customer. Once you trust your brief and your checking, you can move to work that leaves the building, still with a read-through before it goes.

8. Split big tasks into the repeatable part and the human part

Many tasks that look too complex are only partly complex. Writing a client proposal needs your judgement on the offer, but the formatting, the standard sections and the first draft of the summary may be repeatable. Hand over the repeatable part and keep the decision.

9. Keep a "second time" list

The next time you catch yourself doing something for the second time, write it on a list. Do not act on it yet. After a month, the items that appeared three or four times are real repeating tasks. This avoids the trap of building a hand-off for something that only looked recurring.

10. Hand over one job and write it down

Pick one job and write the brief in plain words: the job, the inputs, what good looks like, what to avoid, and a request to ask you questions first. Keeping the brief means you can reuse it. If you use a tool that works with files on your computer, such as Claude Code, which its documentation describes as able to edit files and run commands, the brief works the same way. For help on the brief, read How to Brief AI Like a Senior Hire.

Which two to start with

Not every tactic suits every person. Use this to choose.

If your problem isTry tactics
I do not know where my time goes1, 2 and 9
I have ideas but cannot choose4, 5 and 6
I am nervous about mistakes6 and 7
My tasks feel too big8 and 4
I want to start today10

A short worked example

This is an illustration, not a real case. Say you run a small bookkeeping practice. You read your sent folder (tactic 1) and notice you write a monthly "your documents are due" reminder to clients, tweaking names and dates each time. You score it: it repeats monthly (2), takes 40 minutes in all (3), and is easy to hand off (3). That totals 8. The risk is low if you read each message before it goes. You write a one-paragraph brief with last month's email as the example, run it on last month's client list, compare the result to what you actually sent, and edit the brief where it differs. Whether it saves you time is something you only learn by doing it and checking.

What to avoid

  • Picking the task you dislike most instead of the one that scores best.
  • Handing over a task you cannot describe.
  • Skipping the check because the output looks confident. AI tools can make mistakes, and vendors say so on their own pages.
  • Starting with five jobs at once.

Questions people ask

Do I need all ten tactics?

No. Two are enough to start. Use the table above to pick the ones that fit your situation.

How do I know a task is repeatable enough?

If you have done it at least three times and each time followed roughly the same steps, it is repeatable enough to try.

What if I cannot find any suitable task?

Spend a week on tactic 9, the "second time" list. Most people find several repeating tasks they had stopped noticing.

Should I tell clients I am using AI?

That is your call and depends on your field and contracts. Check any rules that apply to your business, and read each output before it goes out.

Where to go next

Choose your two tactics, run them this week, and write down what you find. When you want a guided path through the same process, Lesson 1.5 walks you through it step by step: Find the jobs to hand to AI. The Find framework turns it into a weekly routine.