AXHUB LECTURE · INTRO L1

Start with one task you repeated today
— the first 30 minutes

Time about 30 minutesWhat you need one free chatbot (ChatGPT, Gemini, or Claude — any)For anyone who has never used AI for work

Doing it once is faster than sitting through ten lectures. This lecture isn't an explanation but a practice. Thirty minutes from now, one of your repeated tasks will actually be handed off to AI.

1Find the repetition (5 min)

Write down five things you did today.

Mark the one that made you think "here we go again." A task with clear inputs, repeated much the same way each time, is your first target.

Don't pick a task that hinges on judgment (evaluating, deciding, negotiating). That comes later.

Practice 1
Things I did today: (1) ____ (2) ____ (3) ____ (4) ____ (5) ____
The repeated one: ____

Common candidates — cleaning up meeting notes, replying to email, answering reviews, sorting inquiries, weekly reports, summarizing material.

Check: is the task you chose "repeated every week (or day)" and "clear on what goes in and what should come out"?

2Write down the procedure (5 min)

Pretend you're handing the task off to a new hire, and write out how to do it in three or four lines.

This note becomes your instruction, as-is.

As you write it, for the first time it becomes clear how you were actually doing this task. That alone makes it worthwhile.

Practice 2 — e.g. if it's answering reviews
These are our shop's rules for replying to reviews.
(1) Thank them (2) Pick up specifically on the menu item or situation the review mentions (3) One line inviting them back.
For a negative review, apologize first, no excuses. Tone polite, two or three sentences.

Rewrite it for your own task. Three rules + one thing to avoid + one line on tone is enough.

Check: could a new hire do the task from this note alone?

3Try it on real material (10 min)

Not a practice example — put in the material you actually received yesterday.

Put in about three of them, so the variation shows.

This step goes as far as finding where something that worked on a sample breaks on the real thing.

Practice 3
[paste the rules you wrote in step 2]

Below are three reviews (email/transcript) I actually received yesterday. Write a draft for each.
[paste the real material]

Even if you don't like the result, don't delete it — it's material for the next step.

Check: how many of the three are at a "good to send as-is" level?

4Have it revise (5 min)

Don't send the draft out as-is.

Check numbers, names, and dates against the original. If the tone is off, show an example and have it fix it.

It's like feedback to a new hire — two or three rounds and it visibly improves. In an analysis of 51 successful cases too, this structure — a person checking at the end — split the results by double (median gain 71% vs 30%).

Practice 4 — feedback examples
Draft No.2 is too stiff. Redo it in about this tone: "Thanks for leaving a note. We'll pay closer attention next time."
And from now on, when an amount comes up, don't calculate it — just mark [amount needs checking].

Check: is there even one sentence you sent out without checking the original (if so, do it again)?

5Save it (5 min)

Paste the instruction that worked into a notes app. When the same task comes tomorrow, use it again.

Within a week it's second nature. Then move to your second repeated task — one a week.

Practice 5 — save format (three lines)
Task: ____
Instruction: [the final version refined in step 3]
Caution: ____ (e.g. always check amounts against the original)

Check: tomorrow morning, could you reproduce today's work from this note alone?

When it doesn't work

The result is flat and genericThat's because the instruction has no standard. Paste in one result you think is well-written and say "this tone and length is the standard." One example is faster than ten lines of explanation.
The result comes out different every timeFix the format — "as a table," "in three sentences," "polite." And instead of rewriting on the spot, paste in the instruction you saved in step 5, as-is.
I hesitate to put in company materialIf you hesitate, mask it and put it in — replace names, contacts, and amounts with ○○. If your company has an AI usage policy, check that first. That there's no policy is also an answer, so the record of having asked is worth it.

Next lecture → L2. How to put AI to work — like briefing a new hire · Full contents

Sources: Stanford Enterprise AI Playbook (51 deployments — review-exceptions-only +71% vs approve-each +30%, 61% of successful projects had prior failures) · workflow automation guide (criteria for automation fit) · reviews from solo operators and practitioners. Full sources are in the axhub.net case library. The practice prompts are example lines to follow along.