The Org.AI Foundation
We are writing down what people do for a living —
and marking, item by item, what a machine can now do instead.
- ·
- ·
- ·
This box does not send your question anywhere. It matches a list built into this page.
- Written down
- 1,016 occupations · 19,000 tasks · 1,012 industriesCounted from the O*NET 29.0 release we loaded, and the 1,012 six-digit national industries in the NAICS release we loaded. Both are countable by hand.
- Sourced from
- 17 outside bodiesNamed one at a time under "Where it comes from", each with the date we loaded it. We said 19 before we checked; 17 is what survived.
- Running as a business
- Not printed yet.We are auditing our own count and the number is not final. When it survives the audit it goes here, and not before. We would rather print nothing than print a number we cannot show our working for.
Work
If it is a job, a task inside that job, the team it sits in, or how you learn to do it — it is in here.
A dental hygienist wonders whether AI is coming for her job.
She types "dental hygienist" into the box at the top of this page. One page comes back — never a list of results. On it are the 23 tasks her job is actually made of, and each one carries a plain badge: a machine already does this as well as an expert · not close · a licence forbids it. Scaling and polishing: not close. Charting and insurance pre-auth: a machine already does this. Her answer fits on one screen, and it is this — the hands stay, the paperwork goes. She never sees a menu, a branch name, or the word "graph."
None of these are open yet. This page is the map.
Agents & Identity
If it acts — a person or a machine — or if it is the leash on the thing that acts, it is in here.
A retiree is asked to let an AI assistant book her travel and pay for it.
She does not type a job. She clicks Agents & Identity, and the word she clicked for is Identity. What she gets is one page about her: one assistant allowed to act for her, what kind of assistant it is, a hard ceiling of $400 a month, the exact wall it has to stop at and come ask her, and then every receipt. One page. That is the whole idea.
Business
If money changes hands, or a licence decides who is allowed to do it, it is in here.
A small-town insurance agent asks what is left for her.
Her job is a chain of steps, and this door marks each one: a machine can do this · only a person holding a state licence may do this. The licensed ones are why she still has a business. The rest are what she should hand over, and there is a page that sells her the thing that does them — with the permission slip attached, scoped to exactly those steps and nothing else.
business.org.ai/starting-a-landscaping-business
Tech
If it is made of software, it is in here. That is the whole test.
A café owner is being sold three different CRMs.
She clicks Tech, because Tech is the word she would have used herself. She lands on one page for the whole category — what a CRM is, what the products under it are, and what people in exactly her situation are actually asking about them this month. She never has to learn that the category and the product are two different kinds of page. That is our problem, not hers.
The World
If it would still be here even if software had never been invented, it is in here. Carbon. Austin. A kilowatt-hour. A forklift.
A homeowner is reading a solar quote and does not know what a kilowatt-hour is.
One link later she is on the page for it: the symbol, what it converts to, and what a household actually uses in a month. Nobody sent her to a physics lesson. This door exists because half of the questions people bring are not about software at all, and a map that only knew about software would send her somewhere useless.
Plain English
Everything in this map already exists somewhere, written for a machine. This is the same thing, written for you.
A parent's kid says "an agent did my homework" and she has no idea what that means.
One paragraph, no jargon, on a page whose whole job is one paragraph. Then two links out: what one of these can do for you, and how you keep it on a leash. This is not a glossary buried at position nine of thirty. It is a front door, because for most people it is the first door.
Where it comes from
Somebody else's record, copied exactly and never edited by us — or our own receipt showing how we got a number.
A reporter wants to know whether any of this is real.
She clicks this word, and gets one page she can print: seventeen outside bodies, each with the part of the map it feeds and the date we loaded it — then a second half, labelled our own receipts. The method behind every badge. The exact file we loaded, its licence and its hash. The proof that two records are the same person's job. Nothing that merely felt like evidence is in the room. Then she scrolls back to the top, reads the third counter, and writes that this is the first AI organisation she has covered that published its own denominator.
Every page will carry this line
It is the same five phrases on every page in the map, so you never have to ask where a fact came from — you just read the bottom of the page you are already on.
Five plain phrases. No jargon, no version numbers, no letters and numbers you would have to be taught.
Where a step has not happened, the line shows a visible blank instead of a plausible name. A blank is information.
This is a specimen. The nurse page above is not published yet.
What this is
The Org.AI Foundation writes down what people do for a living, and marks — item by item — what a machine can now do instead.
We publish the working, not just the answer. Every claim names its source. Every number we made shows its method.
We are building it for the person whose job is on the list. Not for engineers.
And the map is starting to be used. The first program is open: law.org.ai — access to justice, with the lawyer kept in. The second, access to care, is being scoped; nothing about it is built, and this page will say so until it is.