The positioning call I kept almost making, over five weeks.

The idea had been sitting there for months before I actually said it out loud to an AI: our coffee subscription was trying to be everything — beginner-friendly, adventurous, gift-worthy — and it was mediocre at all four. I spent one evening in ChatGPT actually working through what a narrower version would look like: single-origin only, direct-trade relationships we could name on the label, no more variety packs. Nothing decided, just genuinely thinking out loud for the first time instead of just feeling vaguely unhappy about it.

A few days later, after sketching out what it would mean for our supplier relationships, I told it to save where I’d actually landed: “Save the decision — we’re repositioning to single-origin, direct-trade only, dropping the variety packs and the beginner blends over the next two subscription cycles.” First time that thought had existed anywhere except my own head and one conversation.

Then I actually talked to our biggest wholesale account, and it turned out their bestselling SKU was one of the beginner blends I was planning to cut. That’s the kind of thing that doesn’t show up when you’re reasoning in the abstract. I went back and told it: “Update this — the wholesale relationship changes the math, their top seller is one of the blends we were dropping.” Not a reversal yet. Just a new fact sitting next to the decision, making it look less settled than it had a few days earlier.

It sat like that for almost two weeks — I didn’t touch the project once, mostly because I didn’t have an answer. When I came back to it, working through the numbers properly this time, the decision changed: keep one beginner blend specifically for wholesale, go fully single-origin everywhere else. I told it to reverse the original all-or-nothing decision and record the new one, with the wholesale conflict as the reason. The original decision didn’t just disappear — it’s still there, marked reversed, with exactly why it changed, which mattered a lot when my business partner asked “wait, didn’t we already decide to cut everything?” a few days later.

That was the actual point where having this somewhere real paid off. My partner hadn’t been in any of the six conversations — she’d heard the short version over dinner once. Instead of re-explaining the whole arc, I pulled up the project’s Simple view and read it to her off my screen: the current decision, the reasoning, the one thing still genuinely unresolved (how to phase out the other beginner blends without a customer backlash). Two minutes, not a rehash of five weeks of half-remembered conversations.

Before we finalized anything with the roaster, I asked the assistant to turn the project into a one-page brief — “turn this into something I can send them, explaining the change and why.” It wrote from what was actually in the project, the current decision and the reasoning behind it, not from some half-remembered version of a conversation from three weeks earlier. I read it over, changed two sentences, and sent it.

None of the actual thinking happened faster because of any of this. What changed is that five weeks and six separate, scattered conversations turned into one project that knew exactly where we’d landed and why — including the part where we’d landed somewhere else first.