Why decisions
Build from clarity
Products move forward through consequential decisions — not through completed tasks or accumulated deliverables. Everything the studio does is organized around that one claim, so it is worth setting out why we think it holds.
Execution is no longer the constraint
Work can be written, designed, coded, and prototyped faster than at any point in the past. Producing another output is no longer the hard part, and a studio that sells output is selling the part that just got cheap.
What did not get cheaper is knowing which output is worth making. As execution becomes abundant, orientation becomes scarce.
Options are not decisions
More options do not create clarity on their own. They often make commitment harder — every additional plausible direction is another reason to wait.
“A computer can never be held accountable. Therefore, a computer must never make a management decision.”
IBM, 1979
Forty-seven years later, AI can produce twenty plausible directions in an afternoon. It still cannot tell you which one your company should commit to, and it cannot be accountable for that choice. Better prediction does not remove consequence from the person who owns the outcome.
So the line falls in a specific place. Tools execute. AI can reduce uncertainty. People decide — and remain accountable for the hill.
When judgment is actually required
Attention is not distributed randomly. We are drawn to moments where uncertainty and consequence are high at once: several futures remain possible, the outcome matters, and someone must choose before certainty is available. Two axes, four territories.
It matters, and the path is known — training, precedent or expertise already removed the doubt. Often the product of enormous accumulated skill: turning what was once uncertain into repeatable practice is what mastery looks like.
Several futures are genuinely possible, no procedure removes the ambiguity, and someone must still choose. Founding or pivoting, positioning, naming, choosing a direction, committing to a strategy.
Predictable and low-stakes: admin, maintenance, the standard checklist. Necessary, and the first thing worth automating — there is little reason to spend scarce judgment here, and finishing it is not progress.
Sketching, prototyping, playing with a possibility while being wrong is still cheap. Worth protecting deliberately: it is where the options come from, and without it decisions get made from the first available idea.
Uncertainty alone creates curiosity, but without consequence it stays play. Consequence alone creates importance, but without uncertainty it becomes execution. Their intersection creates judgment — and that intersection is the only thing we ask a client to buy.
Why the decision is the record
People remember turning points more readily than completed task lists. A story without uncertainty is predictable; a story without consequence is trivial. What gets told is when someone chose, changed direction, refused, committed, or began — actions matter because they express those decisions.
Records of activity — tickets closed, files delivered — show what happened while obscuring why the product changed. The more meaningful unit is the decision: what was uncertain, what was learned, what was chosen, and what that choice made possible.
It also compounds. Expertise moves a problem leftward on the uncertainty axis — the consequence stays high, but experience has produced better models of it. That is what reflection is for. The artifact may solve today’s problem; accumulated judgment helps solve tomorrow’s, so the climber does not repeatedly return to the bottom unchanged.
What follows from it
- Questions before tasks. Tasks feel concrete, which is why most work starts there. Questions determine direction — better questions produce better decisions, and execution without a meaningful question only accelerates movement.
- Decisions are permanent. A change of direction is a new decision that supersedes the old one, never an edit. The link between them is the lifecycle, and it means a reversal stays legible as a reversal.
- Keep the rationale, not just the choice. A decision without its reasoning cannot be re-examined later, only obeyed or ignored. What was uncertain at the time is the part that ages best.
- Do not automate judgment merely because execution can be automated. Use AI to clarify the terrain and generate possibilities. Require a person to accept a decision before it enters the record.
- Do not count what was finished. A hundred routine completions need not outweigh one decision that changed a hill’s direction. Actions create motion; decisions change direction, and only one of those is worth a scoreboard.