Short answer: Managers can use AI in their weekly work by handing it the repeatable, draftable parts of the job - first drafts, summaries, preparing for a hard conversation, laying out options before a decision - and keeping the judgement for themselves. The version that sticks starts with one task you already dislike, not with a tool. This is the practical, small-steps end of what I think of as ai for managers.

Key takeaways

Move	What it means	When it pays off
Start with a task you dislike	Pick one weekly job you put off, and ask if AI can do the boring middle of it	This week
Frame the problem first	Know the real question before you type anything	Every prompt
Decide what you will verify	Check names, numbers and claims against a source you trust	Before anything leaves your desk
Use it as a thinking partner	Let it widen the options, then make the call yourself	Decisions and hard conversations
Expect the first version to wobble	Tweak the workflow once or twice, then it settles	Weeks one and two

Start with the task you dislike

I'm going to start with a confession, because it's the most honest way I know to explain this.

In 2019 I set myself a challenge: 365 videos in 365 days. The first one ran half an hour, rambling, with some swearing. Maybe one person saw it.

I got better at the videos. What I never got better at was the editing.

I didn't enjoy the grind of it. I couldn't sit through my own video, watch it again and again, and cut it down. So I gave up. (Not my proudest moment, but there we are.)

Years later I paid for a $100 subscription and asked one question: if this can make these videos, can it edit them? My guess was that it could pull the transcript and build a time-based edit. The guess was right.

It didn't work perfectly at first. My wiring was a bit messed up, and the workflow needed tweaking before it behaved. But once it did, it was just so smooth.

Here's the part that matters for a manager. I didn't start with a tool. I started with the one job I'd quit because I disliked it. That's a much better filter than "where could we use AI?", which is a question with no edges.

So look at your own week. What do you do every Friday that you quietly resent? The status summary, the meeting notes, the first draft of the same email. That's your starting point.

Frame the problem before you prompt

The second thing I've seen is less about tools than about thinking.

I ran a two-day workshop called "AI as a Thinking Partner" for a Fortune 100 company, across three batches of about 22 professionals each. The brief was interesting, because the gap wasn't access to the tool. It was mental models. People didn't know how to frame a problem before prompting, or how to avoid hallucinations, or how to get a decision they could actually use.

So we spent a lot of the time on framing. We used the Double Diamond, which makes you separate the real problem from the symptom, and Lotus Blossom, which gives you a structured way to generate ideas.

Participants rated the programme 8.82/10 on average, across 113 respondents. In every one of the three batches, Day 2 rated higher than Day 1. Day 2 was the application day, and I think that's the point. People rate the day higher when they leave with something they can use on Monday.

One participant wrote that the most valuable thing was "Framing questions to AI for better outcomes". Which is about as plain as it gets.

For your week, it means a small habit. Before you open the tool, write one sentence: what is the real question here? If you can't, the AI can't either.

Decide what you will trust and what you will verify

AI can be confidently wrong. Everyone says this, and it's still easy to forget at 6pm on a Thursday.

The habit I teach is boring on purpose. Before you use an output, decide what you'll trust and what you'll check. Names, numbers, dates and any claim that someone else will quote back to you all get checked against a source you already trust. The rough structure of a draft, you can usually take or leave.

You can also ask the AI to tell you where it is unsure. It won't always know, but it's a useful nudge.

Use it as a thinking partner, then make the call

The weekly jobs where I think AI helps a manager most are the ones where you're stuck at the start, not the finish. Laying out three ways to approach a problem. Preparing for a difficult conversation by asking what the other person might be worried about. Turning rough notes into a first draft you can then fix.

None of that replaces your decision. It gives you more to decide with.

One more thing from those workshops. The in-person discussion on Day 2 was rated highly, because different people looking at the same problem produced better solutions than anyone using AI alone. So if the decision matters, talk it through with a colleague as well. Do both.

Expect the first version to wobble

My editing setup took tweaking. Yours will too.

The first week, you'll get something that's nearly right. That's normal, and it's useful, because you learn what you actually wanted. Teach it how you write. Give it the words you really use. Then run it again.

I still haven't sorted the audio-quality bit of my own setup, by the way. I'm looking forward to it.

A short practice for one week

Pick one task you dislike and do every week. Write one sentence on the real question or outcome. Ask AI to do the boring middle of it. Check the names, numbers and claims yourself. At the end of the week, write down what broke and what you'd change.

That's it. One task, one week. If it works, take the next one.

Frequently asked questions

Getting started

How can managers use AI in their weekly work?
Hand it the repeatable, draftable parts of the week: first drafts, summaries, preparation for a hard conversation, and laying out options before a decision. Keep the judgement, the relationships and the final call for yourself. Start with one task you already dislike.

Where should a manager start with AI?
Start with one task you dislike and do every week. Ask whether AI can do the boring middle of it, run it for a week, and note what broke. A small, real task teaches you more than a general tour of tools.

Doing it well

How do I stop AI from giving me wrong answers?
Frame the problem clearly before you prompt, and decide in advance what you will trust and what you will verify. Check names, numbers and claims against a source you already trust, and ask the AI to say where it is unsure.

Does using AI mean managers stop thinking for themselves?
No. The most useful way to use it is as a thinking partner. It helps you frame a problem and see options, and you still make the decision. Talking the problem through with colleagues as well gives you perspectives that individual AI use does not.

Where to go next

If you'd like the longer version of where this goes for a team, I've pulled it together under ai for managers.

With love from Bengaluru, this is Arvindh saying over and out.