You may have noticed I went quiet. Two weeks, no edition. I took a proper holiday, the first real time off since this whole thing started and for the first time in months, I was not sitting inside the system, I was looking at it from a beach-towel distance.
Distance turned out to be the one tool I had never tried. Because somewhere in that fortnight, a conclusion arrived. Part one of this journey is finished. Here is what it taught me.
The conclusion
AI is like everything else.
I know how flat that sounds. But after everything, the agents, the notebook, the migration, the audit, it is the truest sentence I can write. AI requires effort, education, and learning from people who are further down the road than you. Without those three, you will spin some wheels. You will build some automation. You will feel busy and look impressive. And you will never actually gain back the time you spent on the problem you were trying to solve.
I am not guessing at this. I built the hundred-and-twenty-table system and the workforce of note-takers to prove it.
If you are one person
Do not shy away from AI.
If you want it to look at your calendar, make suggestions, help with your emails, do some research, brilliant. Use Claude, ChatGPT, Gemini, whatever the business you work for allows. The more you play with it, the more confident you get. It is the cheapest education going.
If you are rolling it out for a team
Different advice entirely.
If you are thinking about multiple automations, different triggers, systems acting on people's behalf, go and learn the foundations of software engineering first. Or go and speak to people who build software for a living.
(This was the most expensive lesson of part one. I built first and asked second, and the asking is what saved me. You will lose less time, you will learn more, and you will come out of it educated rather than just busy.)
One warning
The deeper you go, the more addictive it becomes. Nobody told me that at the start, so I am telling you now.
Where I am now
Committed, is the honest answer.
I am learning software engineering — properly this time, fundamentals first. I have bought a Mac Mini to build and play with local LLMs, and to experiment with non-business data: building databases with my own hands, breaking them, seeing how they actually work without risking anything real.
(A local LLM is an AI model running on your own machine rather than in someone else's cloud. Nothing leaves the house. It is a sandbox where mistakes are free.)
The end of part one
So this is the natural conclusion of the first phase. There is a second phase — it is already underway — and I am more than happy to keep sharing it. Fair warning: it may get a bit more technical, maybe even boring for some.
So here is my ask, and it is the only one I will make: if you want part two, hit reply and tell me. One line is plenty. That will decide it.
To everyone who has commented, replied, and given feedback throughout these editions — THANK YOU. Learning from other people was the whole lesson of part one. You were part of the education!