Too busy to make yourself less busy

My manager, Sue Dancer, said something I suspect a lot of people are feeling: she knows she should be doing “agent things”, but she can’t find the time to get herself up to speed.

I understand that. The work is already filling the day. Finding time to learn a different way of doing it feels like another job.

There’s an awkward loop here. You need time to work out how AI could help, but the reason you want the help is that you don’t have enough time.

And “get up to speed with AI” is a spectacularly unhelpful item to put on your to-do list. Where does it end? What would you tick off?

I’ve been thinking about a more practical way through it. One that starts with the work already on your desk and gives you a small, useful next step.

Start with the working day you want

Take a moment to describe what you’d like to be different.

For me, I want to spend less time driving AI through every step. I want it to work towards a goal, make useful progress, and bring me back in when it needs my judgement. I want to set direction, review the work, and make decisions.

That’s a direction I can work towards incrementally.

Yours might be more specific:

• Have a first draft of your weekly update ready to review.
• Spend less time assembling information before a meeting.
• Turn customer conversations into a useful summary and follow-up list.
• Keep a project moving without manually chasing every loose end.

Describe the change in your working day. It gives you something to aim at and a way to decide whether an experiment has helped.

Keep the vision loose. You’ll learn more once you start.

Pick one recurring frustration

Choose a task you already do regularly.

Something familiar is a good starting point because you know what goes into it, what usually goes wrong, and what a useful result looks like. You also get another chance to improve it next week.

Let’s use a weekly team update.

Perhaps every Friday you gather notes, look through project activity, work out what matters, and write a summary. The writing might be straightforward. Finding and sorting the information takes the time.

Your ambition could be:

By Friday morning, I want a draft update that identifies progress, blockers, and decisions needed, with links back to the source material. I review it before it goes anywhere.

That’s enough direction to begin.

Work out what you can do with what you have

Spend a few minutes checking the practical constraints.

Which AI assistant do you already have access to? What information are you allowed to give it? Can it reach the material you need, or will you have to provide that yourself? Can it perform actions, or can it only prepare something for you to use?

These answers shape your first step.

If your assistant can’t access your project tools, you might start by giving it a set of approved notes. You can still find out whether it produces a useful update before investing time in connecting anything.

You may discover a dependency that needs help from someone else. Put that on the roadmap. Meanwhile, see whether there’s a useful part you can try with the resources already available.

Keep this assessment brief. It should help you choose an experiment.

Turn the ambition into twenty-minute steps

“Automate my weekly update” is still quite a big assignment.

Break it into pieces with a clear finish line:

• Find a previous update you consider good and remove anything you shouldn’t share.
• Ask your assistant to identify its structure and the information needed to produce it.
• Give it a small set of current notes and ask for a draft.
• Compare that draft with what you would have written.
• Adjust the instructions to address the most important mistake.
• Try the revised instructions on the following week’s notes.

Each is a possible twenty-minute session. Some will take less. Others will reveal that the task needs splitting again.

Twenty minutes is a useful constraint because it forces you to make the next action concrete. It isn’t a promise that every problem can be solved before your next meeting.

A session can end with a draft, a reusable instruction, or a decision that saves you from pursuing a poor idea. Finding out that something won’t work with your current tools is progress too.

Keep a personal roadmap

Write the steps down somewhere you’ll return to.

For each one, record the action and what “done” means. Keep the next action obvious.

For the weekly update, an early roadmap might look like this:

Keep later possibilities nearby without letting them crowd out the next step. Connecting a project system or scheduling a draft can wait until you know the draft is worth producing.

The roadmap will change as you learn. That’s part of its job.

Let your assistant help you make the change

Use an ongoing conversation with an AI assistant to work through the roadmap.

Explain your desired outcome, the resources you have, and the constraints you’re working within. Ask it to help break down the work and tackle one step at a time.

When you return, you should be able to say:

I have twenty minutes. Help me make the next useful improvement.

Keep a short, current roadmap in the conversation: the goal, what you’ve tried, what worked, what’s blocked, and what comes next. Save that summary somewhere you control so you can restart if the conversation becomes unwieldy or you change tools.

An ongoing conversation gives you somewhere to continue. Background progress depends on the tools and setup you actually have. Initially, you may be doing the starting, supplying the information, and reviewing every result.

That’s a perfectly workable place to begin.

As the process becomes dependable, you can decide which additional steps to hand over.

Check whether it helps

After a few attempts, look at the whole effort.

How long did you spend preparing the information? How much checking did the result need? What did you have to correct? Was the final output better?

A draft that takes seconds to generate but twenty minutes to repair may not be saving you anything. It might still be useful if it improves quality, but be clear about the benefit.

For the weekly update, you could keep a simple note of the time spent and the corrections required across a few runs. That gives you a basis for deciding whether to continue.

Keep what helps. Adjust what almost helps. Drop what costs more than it returns.

When you recover a little time, consider putting some of it towards the next improvement. That’s how the roadmap starts creating room for itself.

Give your team a useful starting point

Once an experiment works, share enough for someone else to try it.

Include the task, the instructions, a suitable example, the benefit you observed, and what still needs checking.

“I used this to draft my last three weekly updates, and these are the two things I still have to correct” gives a colleague a practical starting point.

Let people adapt the approach to their own work. Capture those improvements somewhere the team can find them.

If you manage the team, make room for the experimentation inside the working week. A twenty-minute task still needs twenty minutes. Give people permission to use that time and recognise useful learning, including experiments that don’t pan out.

Over time, those individual roadmaps can reveal shared needs. Several people might be assembling the same information or struggling with the same handover. That’s a useful signal about where a team investment could help.

Your first twenty minutes

Choose one task you did this week that you’ll have to do again.

Open your assistant. Describe the task, what makes it frustrating, and what you wish happened instead.

Then ask it to help you find one small improvement you can test today.

I still want to get to a place where AI can make more progress towards my goals with my oversight. The route there starts with making one familiar piece of work easier, learning from it, and returning for the next step.

The prompt below is a place to start.