10 ways to use AI at work this week

Why start with small tasks

The 10 tasks below need no new tool and no setup: all of them fit inside the chat you already open in the morning. What changes from one to the next is the request — and the context you paste in with it. For the bigger picture without the hype, the Stanford AI Index Report is the reference worth reading.

The 10 ways, one by one

1. Rewrite a difficult email

Paste your draft and say: “Make this email more direct and professional, without sounding aggressive.” For a sensitive message (a delay, a chase, a no), ask for two versions — firm and softer — and pick the one you'd actually send.

2. Build a meeting agenda

“Build a 30-minute meeting agenda about [topic]. Include the goal, 3 discussion items, and suggested time per item.” The time limit forces a decision about what stays out — the part a hand-written agenda usually skips.

3. Summarize a long document

Paste the text and ask: “Summarize in 5 bullet points, highlighting the decision items.” Works for reports, contracts, and slide decks. Then flip it: “List what's missing that I should ask about.”

4. Prepare 1:1 questions

“I manage someone who's been on the team for 6 months. Suggest 5 questions for a development 1:1.” Toss the two most obvious ones: the value is in the questions you wouldn't have thought of alone.

5. Turn notes into slides

Paste your notes and ask: “Turn these into 5 slides, each with a title and 3 bullets.” Then ask for the deck's one-sentence takeaway — if it doesn't exist, the slides are tidy but not communicating.

6. Generate title options

“I need 10 titles for an article about [topic]. Professional tone, max 8 words.” Delete the first seven without guilt: the list's job is to show paths you hadn't imagined, not to hand you the final headline.

7. Proofread a text

“Proofread this text: fix grammar, improve clarity, and suggest cuts.” Ask it to explain the single most important correction in each paragraph — that is the difference between shipping a text and writing the next one faster.

8. Compare options

“Compare options A and B in a table with pros, cons, and final recommendation. Use max 5 criteria.” The cap keeps the table readable. Set criteria before pasting the options: decide after the table and you inherit its bias.

9. Create a checklist

“Build a 10-item checklist for [task]. I want to check things off as I go.” Edit it afterwards: AI over-collects. Cut anything whose absence wouldn't matter and the list becomes useful again.

10. Write a spec

“Write a one-page spec for [feature]. Include: goal, scope, out of scope, and success criteria.” The “out of scope” line prevents the argument later — and it's the first thing cut when a spec is written in a hurry.

The secret across all ten

None of these work well without specific context. “Help me with an email” doesn't cut it. “Help me with an email to client X about the delay on project Y” does.

It's the same logic as the AI that asks before answering: the closer the request feels like talking to a colleague who knows the project, the fewer correction rounds you burn.

How to pick your first three

Do not try all ten in one week. Start here:

• If your calendar is meetings: start with the agenda (2) and the 1:1 questions (4).
• If you write all day: start with the hard email (1) and the proofread (7).
• If you decide all day: start with the option comparison (8) and the spec (10).

Three are enough to change a week. The rest follow once the first habit sticks.

Frequently asked questions

Do I have to pay for a tool to do this?

No. The free tier of any major model covers every task here. Paying only matters once daily volume gets in the way — stay on free until then.

Can I paste company data?

Check your account's privacy settings and the provider's policy first. When unsure, anonymise: a client name becomes “client A”, a number becomes a range. Sensitive data rarely buys enough context to be worth it.

Are the answers good on the first try?

Rarely — and that is not the point. The first reply exists so you can correct direction: fix tone, format, and length in short rounds, not by rewriting the request.

How long does this take per day?

Minutes, once the request feels familiar. The gain comes from cutting back-and-forth, not typing faster — and learning time pays itself back quickly.

Follow the path from beginner to pro. And for the dead time outside work, see AI in personal life.