Your first 30 days

What you'll be able to do afterwards: Have one working workflow, one measured result, and a team that knows the rules.

Four weeks, four steps, one workflow. The point is not to transform anything by Friday. It is to finish the month with one thing that demonstrably works and a team that knows the rules.

Everything here assumes you have read the previous nine lessons. If you have not, lesson 1 and lesson 5 are the ones to go back for.

Week 1: Watch, and pick one

Do not buy anything yet. Spend this week observing.

Write down your ten most time-consuming recurring tasks. For each, note roughly how long it takes and how often it happens. Be specific about the time — this is your baseline, and you cannot reconstruct it later.

Mark each one against the five from lesson 3. Drafting, summarising, reformatting, sorting, extracting. Anything marked "none" is not a candidate this month.

Pick exactly one. The best first task is repetitive, bounded, checkable and low-consequence. If two qualify equally, pick the more boring one.

Write down the current numbers. Time per instance, and how many instances per week. Five timed runs is enough. Store it somewhere you will find it in four weeks.

One task. Resist adding a second, even if you can see five more.

Week 2: Build it, by hand first

Run the whole task manually with the assistant, doing each step yourself. No automation yet. You are checking that the output is actually good enough to use, which is the question that decides everything else.

Write your briefs properly using the four parts from lesson 2 — task, context, format, example. Save them. These are the assets you will reuse for years.

Apply the three questions from lesson 5 to anything the model produces that contains a number, a date or a name. Get into the habit while the stakes are low.

Then write the review gate. One sentence: what does a person check, and what makes them stop? If you cannot write it clearly, the task is not ready to automate.

By the end of this week you should have one task that produces usable output with a defined check. That is a real milestone, even though nothing is automated yet.

Week 3: Automate it, in order

Build in the order from lesson 7: extraction, then drafting, then approval, then the final action. One step at a time, testing after each.

Set a spending ceiling in the automation tool before you switch it on. Not after.

Run it alongside the old way. Both processes, same week. You are looking for correction frequency and whether the review time is genuinely lower.

Write down the failure modes — what happens if it stops, if it produces nonsense, if it double-runs — and what catches each.

Expect this week to be the frustrating one. That is normal. Most of the work in automation is the exceptions, not the happy path.

Week 4: Measure, decide, and set the rules

Re-measure. Same five instances, same timer. Compare with week 1.

Decide honestly. Keep, improve, or stop. All three are legitimate outcomes, and stopping quickly is how you find the one that works.

Write your one-line data rule from lesson 8, and check the training setting on your main tool. Two minutes each.

Tell the team what you learned. Not the tool — the actual finding. "This saved twelve minutes per enquiry and needed one correction in twenty." That is what makes people try it themselves, far more than a recommendation.

Set a quarterly date to review the subscription list. This is the most reliable way to stop the slow accumulation that eats the savings.

What to do in month two

If month one worked: pick a second task of the same shape, and keep the first one running. Two working workflows is a genuinely good position at sixty days, and it is more than most organisations have.

If month one did not work: you spent four weeks and learned that this particular task was a poor fit. That is a cheap lesson, and you now know what a good fit looks like. Pick again with better aim.

What you should not do in either case is write a strategy document. The organisations that get value here are not the ones with the best plan. They are the ones that ship one small thing, measure it, and do it again.

The habit underneath all of it

Automate the doing. Keep the deciding and the signing. Measure honestly.

That sentence will get you further than any tool you could buy this week.

This is the end of the series. If something here did not land, or you found something that works differently in practice, leave a comment — the track gets revised when readers say so.