Outcome Coach
Write the outcome — any input. A rough thought, an objective, a whole OKR, the sentence your exec team argued about this morning, or a photo of the whiteboard they argued at. You’ll get a score against the outcome definition principles, the reasons, the questions you most need to answer — and once you have, help writing a better one.
Zero data retentionYour draft goes to a model that keeps nothing and never trains on it — and no account, no database, no copy here. How that works.
Walk the column → The shell the next version of the coaching runs inside: one page from hello to the canvas, with nothing replaced on the way — the site’s front door now. Being built in the open, a card at a time.
How the coach works, and what it can’t do
Your draft goes to an AI model, holding the community’s principles as its brief, and comes back scored against the seven outcome checks — 2 for strong, 1 for partly there, 0 for missing, out of 14 — with the reasons in your own words, the questions that would move the thinking furthest, and candidate phrasings to make your own. It follows the coach design principles: it asks rather than tells, and it never invents a baseline, a target or a fact you didn’t give it — anything it doesn’t know is left in «guillemets» for you to fill in.
It asks before it writes. On a first draft you get the score and two or three clarifying questions — who you are in this, who the outcome is for, what you hope changes for them — and no suggested wording yet, because a phrasing built on the coach’s guesses about your organisation teaches you nothing. Answer what you can and the wording comes next; rough, rounded and anonymised answers are enough, and none of the questions need a commercially confidential number. If you’d rather not answer, skip them and it will offer wording with «placeholders» instead.
It can still be wrong. It reads a sentence, not your organisation, so it can miss context your team would take for granted, and it can be charmed by a bad goal in good vocabulary. Treat the score as a prompt for a conversation, never a verdict. You stay accountable for the goal.
On privacy: your draft and your answers are sent to this site and on to an AI model, through Vercel’s AI Gateway, to write the review. Nothing is stored here — there is no database and no account — and the conversation lives only on this page until you leave it. Every call is made under zero data retention and no prompt training: the gateway will only route it to a provider that has agreed to keep nothing once the reply is written and never train on it, and if no such provider can serve the model it refuses the request rather than sending your words somewhere that would keep them. Web requests to this site are still logged by the host, so anonymise anything sensitive before you paste it. The coach never needs to know who anyone is, and if personal information turns up it sets it aside and works with the role. More on the privacy page.
Pictures are different, and deliberately so. When you hand it a photo or a screenshot, the text recognition runs inside your own browser: the image is never uploaded, never reaches this site, and is forgotten when you close the tab. Only the words it finds go into the box, where you can edit or delete them before anything is submitted — so a face, a name or a colleague’s handwriting caught in the corner of a whiteboard shot never leaves your device. The first picture you read downloads the recognition engine to your browser, which takes a few seconds.
Did it get your goal wrong? That is the most useful thing you can tell us, and it is how the coach improves. Say what it missed — ninety seconds, anonymous if you like.