Outside-in editions.
The instrument, run against real organizations that never asked. Each one is built from the public record alone - annual filings, Form 990s, a live job posting, their own front door - and nothing else. If the method works with no inside information, consider what it does with yours.
The standing rules.
They are the same for every edition, and they are the reason these are worth reading rather than simply being about someone real.
Public sources only
Filings, 990s, live postings, public web. Nothing non-public is used or sought, and each edition says so on its face.
Every claim carries its source
Linked at the claim, not gathered in a bibliography nobody opens. Quotations are checked against the document they are attributed to before an edition ships.
What the record does not show is said, not filled
An outside-in reading has limits. Where the public record goes quiet, the edition says so rather than inferring something plausible.
No headcount-reduction recommendations, ever
The same rail the instrument runs under. It holds in the editions too.
Where the same work gets filed.
Each of these organizations is hiring a senior person to lead AI strategy. Each classified the role on its own careers board, in a dropdown, before anyone was hired and without meaning to say anything.
Where an organization files the work is a statement about what it believes the work is. It is also the cleanest example in this library of the distinction the ceremony audit runs on, and nobody had to be interviewed to find it.
| Organization | The role | Filed under |
|---|---|---|
| Charles Schwab | Director, Responsible AI | Risk & Regulatory |
| Applied Materials | Senior Director, Enterprise AI & Automation | Information Technology |
| Novartis | ED, NovaOS Strategy and Innovation - AI | Marketing |
Each taken from the employer's own posting on the day it was read. One of these organizations is going to be surprised by what it hired.
What the instrument returns.
Harbour Freight Lines
Sample data · the QUICK path
A sample Kurultai document - written by the live engine, from an invented company.
Every other edition on this page is outside-in: the instrument reads an organization from its own public record, and the organization never typed anything. This one is the ordinary path, and the only one that shows what a visitor actually gets back. Harbour Freight Lines is a 400-person regional logistics company that does not exist; its four positions and four paragraphs went through the same engine, the same rails and the same cut line as any other run, and nothing in the document was edited afterwards. It is the fixture the QA harness runs, so the thing demonstrated here and the thing tested are one cast.
A demonstration, not advice. The document says so at the top of itself.
The library.
Charles Schwab
Public filer · 10-K path
Charles Schwab's Kurultai - read against the FY2025 Form 10-K, the 2026 proxy, and their own careers board.
The edition where the instrument finds the least to correct, which is why it is worth reading. Schwab split the AI leadership role into the five functions most organizations smush into one job description, and staffed them separately. The finding is what their shareholder record does not say about it.
Applied Materials
Public filer · 10-K path
Applied Materials' Kurultai - read against the FY2025 Form 10-K, the 2026 proxy, and their own careers board.
The counterpart to Schwab, and the library's thesis exhibit. Their filings name AI a top-tier enterprise risk with board oversight; the operational answer is one seat holding five named responsibility blocks, filed under Information Technology, reporting to a VP of Engineering, with responsible-AI governance listed as a preferred qualification. They are paying less for five functions than Schwab pays for one.
Novartis
Foreign private issuer · 20-F path
Novartis' Kurultai - read against the FY2025 Form 20-F and their own careers board. A third document regime: no 10-K, no proxy, Swiss disclosure.
The best-shaped AI leadership posting in the library - a strategist asked to do strategy, with delivery explicitly elsewhere - at the organization whose annual report never once names the platform the job exists to run. Its declared AI risk is also the inverse of everyone else's: not what AI might do to them, but failing to adopt it fast enough.
The Jed Foundation
Nonprofit · 990 path
The Jed Foundation's Kurultai - read against one job posting in full, thirteen years of Form 990s, and their public web presence.
A well-scoped hire pointed entirely outward, at an organization whose inward questions - its own data, its own adoption, its own staff - have no owner in the public record. The 990 path shows less than the 10-K path, and this edition says where it goes quiet.
Each edition is dated and reflects the public record on the day it was read. A filing supersedes an edition, never the other way round.