Your twenty-minute routine

Five lessons, five separate techniques. On their own each is a thing you might remember to do. Assembled into a routine, they become the thing you actually do, which is the only version that matters.

This lesson is short because the routine should be.

Batch, don’t sprinkle

The instinct is to reach for an assistant whenever a task appears. That maximises switching: a different window, a different frame of mind, a pause while you write the request.

Batching works better. Two blocks with a clear purpose:

Morning, ten minutes — orient. Overnight email triaged into what needs a reply from you and what does not. The one long document you have to get through today mapped, with quotes, so you know which sections to read (L3). Yesterday’s notes turned into decisions, actions and an unclear list (L5).

End of day, ten minutes — close. Routine replies drafted in your voice and checked (L2). Tomorrow’s most awkward task decomposed, no timings (L6). Anything you researched today spot-checked against its citations (L4).

Twenty minutes. Not a transformation — a routine.

Make it survive a bad week

The most common failure is not choosing the wrong tasks. It is building something that only works on a good day.

A routine gets tested in the week you are ill, travelling and behind. If it needs forty minutes and a clear head, it will not run, and a routine abandoned in week two has saved you nothing while costing you the setup.

So size it for your worst realistic day and let it be unimpressive. You can always do more on a good day. You cannot retroactively do a routine you skipped.

Keep the receipts

For the first fortnight, note two things: what you handed over, and whether checking it took longer than expected.

That second column is the one that matters. It is L1’s verification cost, measured rather than estimated, and it is the only honest way to find out which steps are earning their place. Some will surprise you in both directions.

Then cut. Any step where checking reliably takes longer than doing the task yourself comes out — not because the assistant did badly, but because that step was never a saving.

What this does not do

Worth saying plainly, because a lot of writing about this promises otherwise.

This routine will not double your output. It removes friction from a specific set of frequent, low-stakes, easily-checked tasks — the first box from L1. The hard parts of your job, the ones requiring judgement, relationships and responsibility, are untouched. That is the subject of the next lesson.

Try it now (6 minutes)

Write your two blocks on one line each, using only tasks you already do.

Run it tomorrow. Then run it the day after. If you skip a day, make the routine smaller rather than resolving to try harder — the size was the problem.

Check your understanding

1. Why batch AI-assisted tasks into blocks? Assistants respond faster in the morning Switching tools has a cost, and two purposeful blocks beat twenty interruptions It uses fewer messages on a free plan

2. How large should a new routine be? As large as your best day allows Small enough to complete on your worst realistic day It does not matter, as long as it is written down

3. Which step should be cut from a routine first? Any step where checking takes longer than doing it yourself The longest step The step you enjoy least

Recap

Two ten-minute blocks — orient in the morning, close at the end of the day — built from tasks you already do. Size it for your worst day, track what checking actually costs for a fortnight, and cut the steps that do not pay. It removes friction from routine work; it does not touch the hard parts.