Treat AI like a brilliant chief of staff with no memory of you. Capable. Fast. Has no idea what you want or how you decide. This module is foundational — everything you'll build on top compounds from this one skill.
Left to its defaults, AI validates. It fills gaps in your reasoning with plausible-sounding content. It produces polished output that looks right even when the thinking underneath is soft. This is not a flaw to complain about — it is the architecture.
The question isn't whether AI will agree with you. By default, it will. It's the new chief of staff trying to look competent. The question is whether you've briefed it well enough — and pushed back hard enough — that the agreement is worth anything.
Cognitive ease is the tell. When your chief of staff makes a hard problem feel easy, they've either solved it or skipped it. Usually skipped it.
Journalist Zach Seward described effective AI workflows as "humans first and humans last, with a little bit of powerful, generative AI in the middle to make the difference." Three operational constraints.
Brief your chief of staff like they're new — because they are. What's the problem? What counts as evidence? What does "good" look like? Skip this and you'll get confident work that solves the wrong problem. Framing is the executive act. Never delegate it.
Now use your chief of staff against you. Surface assumptions. Force counterarguments. Make it argue the other side. A great chief of staff disagrees with you on purpose. Your AI won't — unless you tell it to. If the work feels easy, you're being agreed with. Push harder.
You make the call. Then have AI write the decision record — the call, the options you rejected, the assumptions you're betting on, the indicators that would tell you you're wrong. You'd never write this by hand. AI will. Log it where future-you can find it. Your chief of staff keeps the paper trail. You keep the judgment.
If AI makes your problem feel instantly resolved, you are likely skipping a step that protects your judgment. Slow down and ask harder questions.
Two skills compound here: scoping (getting precise about what you actually want) and learning (using AI's pushback to sharpen your own thinking). Every loop sharpens both.
The routine below is built for the moments when the point is not to generate content, but to reach a defensible decision without degrading judgment. Deliberately repetitive. That's what makes it work.
Write 150 to 250 words answering: What's the problem? Why now? What decision is required? No AI. Sometimes you'll know the problem and this is your briefing — to yourself first, then to the chief of staff. Sometimes you won't, and the act of writing is how you scope it.
Either way, the ugly first paragraph is diagnostic. It reveals what you actually understand and what you're hand-waving past. Then hand it over: identify ambiguous terms, list hidden assumptions, generate three rival framings.
This turns AI into a conceptual editor rather than a ghost writer. You own the framing. AI stress-tests it.
Ask AI to make the strongest case for Option A using only your stated assumptions, then the strongest counterargument. Repeat for each option you are considering. Reject generic objections like "execution risk." Require falsifiable claims — what would you actually observe if this objection is correct?
Example promptsObserve, Orient, Decide, Act. The most common executive failure mode is jumping from data to decision while skipping orientation — the sensemaking step. Use AI to help you separate facts from conjectures, label confidence levels, and identify the assumptions driving each option.
If AI cannot do this clearly, you have learned something about your own evidence base.
Before finalizing any significant decision, ask AI to take the perspective of people who'll challenge it. Convert the output into a pre-mortem paragraph in your own words: what's the most likely cause of failure, and what early indicators would warn you? A chief of staff who only nods is useless. So is an AI you haven't pushed.
Example promptsYou won't write a decision record by hand. Almost nobody does. So have AI write it for you. A one-pager:
Save it where the system can find it later — your project folder, your second brain, wherever your AI lives. Then turn the model on its own work: find missing assumptions, false dichotomies, unsupported claims. Fix them yourself.
The decision record is the artifact your chief of staff produces while you do the thinking. The point isn't a clean doc. It's that next quarter, you (or your AI) can pull it up and check the math against what actually happened.
Thinking is a perishable skill. Deliberate nonuse is as important as deliberate use. A simple maintenance cadence:
Choose what you offload. Keep practicing what you cannot afford to lose.
Read the full article and the companion piece, The Chat Trap: When AI Makes Your Thinking Softer, at robert.winter.ink. Both are free with a basic account.
Everything coming next runs on the muscle you build here.