The Kurultai
A council for
what comes next.
The Kurultai is a free AI strategy tool from Prodigal Paradigm. Tell it what your organization actually does all day and who actually does it, and it returns a strategy document: four considered positions and a short brief in, a ceremony audit, a decision ledger, a people arc and a staffing map out. It does not produce headcount-reduction recommendations, ever.
Try the instrument QUICK is free.
The AI strategy tool that won't write your layoff plan.
No maturity score. No right answer. Put each marker where your organization actually is.
These are positions, not predictions. The sentences are the scale. The number is only a location.
This takes a few minutes. Leave the tab open.
While you wait — five things worth knowing about the document being written.
A ceremony audit sorts your processes into work judged by what it produces and work judged by how it is performed. A decision ledger turns every freed hour into a choice with a name and a date against it. A people arc runs from MyRunbook, where each person writes down what they actually do all day, to MyWork, where they and their manager decide what the job is now. A staffing map reads your people against five archetypes.
This is the part most strategy documents skip. Work gets faster, nobody decides where the time goes, and it is quietly absorbed — the same work, at the same pace, with the slack gone. So the ledger will not let a row stand without a named person and a date. A row with no owner is not a decision, it is a hope.
Sorghaghtani Beki cultivated political alliances, supported education, and helped bring Möngke to power in 1251, as described in Timothy May's biography for EBSCO. She never took the throne and raised four rulers. Our reading is a design principle: build the conditions for sound decisions, and leave the mythology out. Why the Kurultai
Not at any slider setting, and not if you ask it directly. The rule is enforced twice: once in the engine's own instructions, and again by a separate check that reads the finished draft and stops it before it reaches you. The document may name what you asked for — it is allowed to tell you it declined — but it will not argue for fewer people. What the work frees up is a decision, and the ledger says whose.
This is a sketch built from four positions and a few paragraphs, written by a model that has never met your organization. It is meant to be marked up, contradicted, and handed to the person who knows why every exception exists. Every recommendation names the future outcome it controls, so you can test each one against that claim. The last section lists what the document knows it does not know. Start there.
These cards say nothing about how far along your document is. The engine writes it in one pass, so there is no honest way to report progress — only to tell you when it is done.
Not a use-case generator. Four readings of the same organization, each of which has to survive a second reader.
Judged only by whether it controls a future outcome. That is the bar the audit puts your processes through, so it is the bar the document has to meet itself: every recommendation names the outcome it controls and the evidence that would show it.
Read a full sample documentThe Kurultai · Engine one
Point an instrumental tool at a ceremonial process and the process absorbs it and produces more ceremony, faster.
Every process you describe is sorted into two piles. Instrumental: it is judged by whether it controls a future outcome. Ceremonial: it runs on precedent, or it displays status.
Recommendations attach only to the first pile. The second pile is named for the management conversation it deserves, with no tooling recommendation attached to it.
The Kurultai · Engine two
Freed time stays free only by explicit management decision.
The washing machine gave households back hours and the hours went straight back into the same work, because nobody decided otherwise. The same thing is happening now, faster.
| Capacity freed | Owned by | By when |
|---|---|---|
| What the work gives back | A named role | A date |
A row with no owner is not a decision, it is a hope. The table is meant to be signed.
The Kurultai · Engine three
The piece nobody is selling.
Each person documents what they actually do all day. It is the only way to see what hands off and what cannot, and it is the thing that stops domain knowledge walking out of the building unrecorded.
Having seen it, work out what the job is now. Then the ladder: equipped, oriented, builder, ready to write code with an agent. That is what engaged has to mean if it is going to mean anything.
The Kurultai · Engine four
Five archetypes, mixed for the stage you are actually at.
Prototyper, Builder, Sweeper, Grower, Maintainer. Before product-market fit the weight sits on the first three; growing, it moves to the middle; once the thing is working, to the last three.
Which archetypes are already in the building, which you would be hiring or partnering for, and the hiring conversation named where hiring actually enters.
Start with QUICK. Add detail or a human conversation when the decision calls for it.
The same four engines, built against your own public filings and documents rather than only your self-report, with every claim carrying a primary source.
A conversation to challenge assumptions and put the strategy to work.
The rest of this page is written the way the work sounds. This part is written the way a question sounds.
A free AI strategy tool from Prodigal Paradigm. You describe your organization's processes in your own words and place yourself on four sliders; it returns a strategy document built from that and nothing else.
Organizations deciding what to do about AI and considering hiring someone to tell them: companies, nonprofits, government bodies. It is aimed at the moment before the ill-scoped AI-strategist hire, not after it.
No. The Kurultai does not produce headcount-reduction recommendations, ever. That promise is enforced twice: in the engine's instructions, and again in code on the way out, where a breach gets one corrective retry and then fails the job rather than returning the document. The test battery behind it includes a buyer who sits down and asks for the layoff plan outright. The enforcement is in the public repository, because "enforced, not hoped" is only checkable if you can read it.
It is sent to Anthropic's API to be read, and comes back as your document. Nothing is stored on our side but a count, so the day's limit can be enforced, and the finished document, which deletes itself within two hours. No account, no tracking, no record you did not take with you.
The QUICK sketch is free and needs no account. The DETAILED document and the advisory session are priced on request, because the scope depends on what you are deciding; write and ask.
The ceremony audit sorts your processes into the ones judged by whether they control a future outcome and the ones that run on precedent, and attaches recommendations only to the first kind. The management-decision ledger assigns every freed hour to a named role with a date. The people arc documents what people actually do before deciding what the job is now. The staffing map places you against five builder archetypes and names where hiring enters. They are described above and shown working in a sample document.
Each person writes down what they actually do all day, in their own words. Not a process diagram drawn by somebody else - the work as the person doing it understands it. It is the only way to see which parts of a job hand off to a machine and which cannot, and it is how knowledge that currently lives in one person's head gets onto a page before that person leaves.
What follows MyRunbook. Once the work is visible on a page, the person and their manager decide what the job is now: what it keeps, what it sheds, and what it gains. MyRunbook is the observation; MyWork is the decision. Doing the second without the first is guessing.
Boris Cherny, Head of Claude Code at Anthropic, proposed five team archetypes that transcend traditional job titles - engineer, designer, product manager - in the age of AI (reviewed November 2, 2026). The staffing map reads your organization against them, because a seat advertised as one title is usually asking for two or three of these at once, and that is the thing worth naming before anybody is hired.
Prototyper. Generates brand-new ideas and churns out many concepts quickly, most of which never ship.
Builder. Takes a prototype or rough idea and rapidly turns it into a production-grade product or infrastructure.
Sweeper. Cleans up the interface, simplifies code and systems, unships failing features, and optimizes performance.
Grower. Takes an existing shipped product and iterates continuously to improve product-market fit.
Maintainer. Owns mature systems and keeps them secure, reliable, fast and efficient as they scale.
The mix moves with your product stage. Before product-market fit the weight sits on Prototyper, Builder and Sweeper; growing, it moves to Builder, Sweeper, Grower and some Maintainer; once the thing is working, to Sweeper, Grower, Maintainer and some Builder. No individual is all five, and a posting that asks for all five is the finding, not the candidate.
A maturity assessment scores you. This does not: there is no score, no grade and no right end of any slider. It also refuses to recommend anything it cannot be judged on, so every recommendation names the future outcome it controls and the evidence that would show it.
Kathleen Bartin, founder of Prodigal Paradigm LLC in Millbrook, New York, who has spent thirty years on technology adoption inside organizations. LinkedIn.
