In partnership with
Issue #15 · Ops & Om · Health of Business
You set up the tool. You wrote the instructions, connected the accounts, got it running. It felt like crossing something off — done, that's handled now.
That's the trap. AI setup isn't a thing you finish. It's a thing you keep. And the gap between those two words is where most people quietly lose the leverage they thought they'd locked in.
I'll be honest about how I learned this: I used to roll my eyes every time I had to go update my instructions again. It felt like the opposite of what AI promised. Wasn't this supposed to take work off my plate? It does — but only if you keep showing up to it. The leverage isn't in the setup. It's in the upkeep.
Here's what we'll cover in this issue:
Why a "finished" AI setup is already out of date
The ten-minute ritual I run at the end of every work session
Why this is a standing job, not a one-time task
Your AI Setup Has an Expiration Date You Can't See
Nothing about your setup is static. You teach it a new preference and the old instructions don't reflect it. A new tool ships that makes half your workflow obsolete. You catch it making the same small mistake twice and realize the instructions never told it otherwise. None of that announces itself. The tool keeps running, a little less useful every week, and you blame the model.
I hit this constantly with my own project instructions. I'd finish a great session, then notice the setup that got me there was already behind — we'd connected something new, or found a redundancy we could now design around, and none of it was written down anywhere the AI would see it next time.
That's not a failure of the tool. That's just the nature of a system that learns. A setup that never changes isn't stable — it's stale.
The Ten-Minute Ritual That Keeps It Sharp
So I built a habit. At the end of a work session, before I close the chat, I do one thing: I paste my current instructions back in and ask the AI to turn the session into an upgrade.
Try this prompt:
Here are my current project instructions: [paste them].
We just finished a work session. Review these instructions
against what actually happened today. Flag anything we should
rewrite — a new preference you learned about me, a tool we
connected, a redundancy or mistake we can now design around.
Then give me a clean, rewritten version I can drop back in.It takes ten minutes. The AI just watched the whole session, so it's the best-positioned thing in the world to tell me what changed — and it surfaces things I'd never have flagged on my own. Then I drop the rewrite back in, and next session starts smarter than this one ended.
The email agent I walk through in the podcast is the clearest proof of this. It isn't good because I built it once. It's good because the rulebook gets a small edit every time it guesses wrong — that's the entire reason it went from asking about everything to handling 90% silently. The setup was the easy part. The maintenance is what made it work.
That rulebook — with the safety guardrails already baked in — is the thing I packaged up for you.
This Isn't a Task. It's a Standing Job.
Here's the part nobody tells you when they sell you on AI. The promise is real — it takes enormous work off your plate. But it quietly hands you a smaller, different job in return: keeping the thing intact. Tending the instructions. Noticing what drifted. Deciding what to update when a new tool lands.
Most people never pick that job up, which is exactly why their AI underperforms while someone else's compounds. The difference between the two is never the tool. It's whether anyone is paying attention to it.
There's a whole discipline hiding in that sentence — what it actually looks like to run AI as an ongoing system instead of a one-time install. That's a bigger conversation than one ritual, and it's coming. For now, the ritual is enough.
Run it once this week and you'll feel the difference in your very next session.
Baldomero Garza — Find me on X, LinkedIn, Instagram, or book a 1:1.
P.S. — If you only do one thing this week: paste your instructions back in and ask what to rewrite. Ten minutes, and your setup stops aging.
Talk to your AI tools the way you'd talk to a colleague.
You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.
Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.
Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.
If you found this issue helpful, please forward it to a friend or colleague who might benefit from it. Sharing is caring!
Did someone forward you this issue? Don't miss out on future insights — subscribe to the Ops & Om newsletter here!