Don’t Repeat Your Decisions

AI made it impossible to ignore how often we end up repeating our micro and macro decisions to it.

Decisions are one of the highest leverage actions we humans take. Where we are today is the compound result of decisions we took at given moments in the past; some deliberate, most on autopilot. And as the stakes go higher, you find yourself making more critical decisions. That is what seniority actually is; less doing, more deciding.

Every decision is a bet

But a decision is not a fact; it is a bet placed with the information you had at that moment. The world keeps moving after you place it. So it becomes even more important to know when a past decision is becoming a problem and needs to be changed. In practice, this rarely happens on time. It comes as a late oversight after a jarring failure. The post-mortem reveals the bet had expired months ago and nobody was watching it. We are decent at making decisions. We are terrible at noticing when they go stale.

The decisions you don’t see

There is another layer that stays hidden. When we do an action manually, we don’t realise how many and what decisions are being taken along the way. Our subconscious treats most of them as facts. Which edge case matters, who should be informed, what “good enough” looks like, what order things should happen in. Others are important decisions that aren’t being considered at all. Defaults we absorbed from a team, a culture, a past failure. None of them announce themselves as decisions. They are simply how we work.

Try to automate that work with AI, and all of it becomes visible at once.

AI is inherently not designed for taking decisions. It goes with the highest probability occurrence, the most likely continuation given the context it has. It also doesn’t share our subconscious. Ask AI to draft an email you would have written yourself, and watch what happens. You correct the tone. You correct what to leave out. You correct when to escalate and when to let go. None of these were in your instructions, because none of them felt like decisions when you did it manually.

So we find ourselves repeating the same micro and macro decisions to AI time and again in every prompt, every correction, every “no, not like that.”

DRY, but for decisions

Programmers solved repetition long ago with a simple DRY (Don’t Repeat Yourself) principle. If the same logic appears in two places, extract it into one and reference it everywhere. Repetition is not just wasteful; it is how versions drift apart and how bugs are born.

Decisions deserve the same principle. Whether you are trying to be more productive with AI personally or speed up an organization, the important work now is building a decision layer. One explicit place where the decisions that matter live. What was decided, why, and under what conditions it should be revisited.

For humans, this layer creates strong alignment and makes the bets visible before they fail, not after. For AI systems, it is the missing subconscious that lets them take the same decisions we would have taken, without us re-deciding everything in every session.

Execution keeps getting cheaper. Deciding well and not repeating yourself while doing it is what remains scarce. The leverage was always in the decisions. AI only made it obvious.