Chapter 28: The 30-Day AI Learning Plan

September 2, 2026

This entry is in the series Get on board with Artificial Intelligence

How this works

Thirty days, twenty to thirty minutes each. Every day uses real work from your actual life: none of it is practice for its own sake.

Three rules that make it work:

Use real material. Low-risk work from your actual life, wherever possible. Practice for its own sake teaches far less than practice with something at stake.

Practice daily rather than in large blocks. Twenty minutes a day usually teaches more than three hours on Sunday.

Missing a day is fine. Continue with the next one instead of restarting or trying to catch up all at once.

If you have read only the Fast Track, everything referenced here is in chapters you have read.

Week One: Use it (Days 1–7)

This week, focus on building the habit. Quality will improve after AI has found a regular place in your day.

Day 1: Set up. Create an account. Fill in custom instructions, including an uncertainty instruction. Check your training and memory settings. (Ch 3)

Day 2: The first hour. Run the four-prompt sequence: establish, transform, the deliberately bad prompt, the properly built one. Notice the gap. (Ch 3)

Day 3: Transform something real. Take your own writing and shorten it, change its tone, and restructure it. Three requests, one piece of text. (Ch 2)

Day 4: Pick three from the fifty. One work, one personal, one frivolous. Do all three today. (Ch 2)

Day 5: Learn to steer. Take today’s worst output and fix it with four short corrections instead of rewriting the prompt. (Ch 4)

Day 6: Make it interview you. Something consequential. Open with: ask me up to six questions before you answer. (Ch 4)

Day 7: Run number fifty. I’m a [role] doing [work]. What are five things AI could help with that most people in my position never try? Circle the surprises. (Ch 2)

Week Two: Trust it correctly (Days 8–14)

The goal this week is calibration. By Friday you will know exactly how much to believe.

Day 8: Manufacture a hallucination. Ask for five academic sources in a field you know. Check every one. (Ch 8)

Day 9: Ask it to confirm the fiction. Take a source you proved does not exist and ask, in the same conversation, whether it is real. (Ch 8, Ch 9)

Day 10: Find the silent assumptions. On any substantial advice you have received: what did you assume about my situation that I didn’t tell you? Count the wrong ones. (Ch 8)

Day 11: Run full TRUST once. A real question in your field, all five steps. Time it. (Ch 9)

Day 12: Prove the cold re-ask. One narrow factual question, three fresh conversations, worded differently. Note where they diverge. (Ch 9)

Day 13: See the flattery. Present the same plan two ways: as yours, then as a colleague’s needing evaluation. Compare. (Ch 6, Ch 11)

Day 14: Write your verification bins. One page: what you never check, spot-check, and fully verify. (Ch 9)

Week Three: Do real work (Days 15–21)

Day 15: Build your context file. Ask to be interviewed, then test the result on a real request. (Ch 22, Project 1)

Day 16: Interrogate a document. A real contract or policy. Obligations with exact quotations. Verify four quotes. (Ch 21)

Day 17: Rescue a spreadsheet. Real messy data. Cleaning plan approved first, then all five verification steps. Check the row count. (Ch 21)

Day 18: Set up a project. Load background documents and instructions for something recurring. Run a request inside and outside it. (Ch 12, Ch 13)

Day 19: Build a reusable prompt. A monthly task, refined until genuinely good, then converted to a template with blanks. Save it. (Ch 22, Project 4)

Day 20: Decompose something. Take a task you would normally one-shot. Four stages, a decision at each. (Ch 11)

Day 21: Run the maker–critic loop. Draft in one conversation; critique in a fresh one with no context. (Ch 11)

Week Four: Make it last (Days 22–30)

Day 22: Learn something with LEARN. A feature you have avoided. Paste the official documentation first. (Ch 17)

Day 23: Prove the L step. Ask a tool about one of its own features from memory, then again with current documentation pasted in. Compare. (Ch 17)

Day 24: Use LEARN on something that isn’t AI. A spreadsheet function, a phone setting, work software you use badly. (Ch 17)

Day 25: Audit your exposure. Read your actual data terms. Review memory. Review connectors. (Ch 24)

Day 26: Write your AI policy. One page, five questions. Household version if relevant. (Ch 24)

Day 27: Agree a family code word. Then write your verification rule for urgent money requests. (Bonus A)

Day 28: Build something. A small working tool you would use. Test it against three known answers and one deliberately wrong input. (Ch 22, Project 6)

Day 29: Assemble the folder. Six files. Fill in what exists, leave headings for the rest. Run your first review. (Ch 26)

Day 30: Take stock. Answer four questions in writing: – Which three uses have actually stuck? – What did the verification exercises teach me about how much to trust this? – What am I now delegating that I should still be doing myself? – What is the one thing I want to be able to do by this time next quarter?

After thirty days

After thirty days, you will have a working habit, calibrated trust, and a small system that keeps functioning as the tools change. Mastery can grow from that foundation.

Three things to keep doing:

The quarterly review. Thirty minutes. Use the system you built.

Adding to the library. Whenever something works well.

LEARN. Whenever something new appears. That is how you stay current without a book.