A Prodigal Paradigm product Outside-in edition
November 1, 2026
The Jed Foundation's
Kurultai.
Built from public sources alone: the full text of one job posting, four years of Form 990s, and JED's public web presence. No inside information was used or sought.
What this document is
The Jed Foundation, the nonprofit that protects emotional health and prevents suicide for teens and young adults, is hiring a Director, AI Strategy & Mental Health. The posting is live on JED's own hiring page with a 14 November 2026 deadline, at $120,000 to $160,000, reporting to the SVP, Youth Strategies and Chief Engagement Officer.
This is what the Kurultai produces from public sources alone. JED brought nothing to the instrument; we read the record.
One difference from the Centene edition is worth stating plainly: the 990 path shows less than the 10-K path. Nonprofits file no risk factors, no compensation philosophy, no management discussion. So this edition leans harder on the one document where JED speaks at length about AI, the posting itself, and says where the public record goes quiet rather than filling the silence.
What the posting gets right, first
Candor requires leading with this. It is a well-scoped posting, and in places a sophisticated one.
It says out loud that the role "does not require the candidate to be an ML engineer." It names the teams the Director will work with, Knowledge & Advising, Policy, Programs and MarComms, "to bring relevant expertise into the work," rather than leaving collaboration to hope.
And one responsibility shows unusual depth. The role must consider not only acute safety failures but "relationships and dependency, connection and belonging, agency, coping, help-seeking, protective factors, developmental differences, and beneficial uses of AI." Most AI-safety job descriptions stop at harm. This one understands that a lonely teenager's relationship with a chatbot is a wellbeing question long before it is an incident report.
The instrument notes good scoping where it finds it. That is what makes its findings elsewhere worth weighing.
Where the public record places them
The outward seat
and the inward one
Every responsibility in the posting faces the field: safety and evaluation collaborations, benchmarking, advising technology companies, developing JED positions and frameworks, partnerships, external representation. Not one line addresses the inward questions. How JED itself adopts AI in its own operations. Who governs JED's own data, at an organization whose work sits adjacent to the most sensitive information young people have. What JED's own staff do as AI reshapes their workflows.
The posting runs to roughly 7,500 words and contains no occurrence of governance, privacy, clinical, protocol, vendor or procurement, and none of supervision or direct reports. The public record names no owner for any of it. This document exists for that half.
A role still
finding its name
The posting's header reads "Director, AI Strategy & Mental Health." Its overview says JED "is seeking a AI Safety & Wellbeing" - a different name, and an unfilled slot, mid-sentence. The location designations disagree the same way: "Fully remote" at the top, and lower down, in the template, "Roles at JED are primarily hybrid and based out of the New York City Tri-State Area."
Read kindly, and it should be read kindly, this is what a genuinely new capability looks like while an organization is still shaping it. Read as a strategist, it is the finding: the role's boundaries are not settled, so the first occupant will settle them, which makes the scoping conversation this document maps the one that decides what JED actually bought.
A band, a specialism,
and no stated plan
The posting asks for roughly seven or more years in AI safety, evaluation or responsible AI, with working knowledge of model behavior, red-teaming, benchmarking and failure modes, and offers $120,000 to $160,000.
What the public record shows is the band and the ask. What it does not show is any stated plan for recruiting and holding that specialism, which is in demand well beyond this sector. This document does not put a number on the gap, because no public source here establishes one. It names the decision instead, because recruitment time, retention risk and the seniority of whoever accepts all turn on it. Mission is a real currency; it should be budgeted deliberately rather than assumed.
A seat opened
on purpose
JED's Forms 990 show a decade of growth and then a turn. Revenue rose from $1.0M in tax year 2011 to $29.8M in 2022, with a surplus every year from 2014 through 2022. Tax year 2023 is the first deficit in ten years: revenue $20,673,662 against expenses $25,788,488, a shortfall of $5,114,826, with revenue down about 31 percent while expenses rose about 38 percent.
One year of a 990 cannot separate a genuine decline from a multi-year grant recognized in 2022, and the tax year 2024 return carries no extracted financials yet, so this document describes the shape and asserts nothing about strain. What the shape does support is this: opening a new director-level AI seat in the year after the first deficit in a decade is a deliberate priority rather than a surplus luxury. Somebody decided this capability was load-bearing.
Engine one: the ceremony audit, outside-in
Instrumental,
clearly
The evaluation mandate. "Apply knowledge of model and system behavior, safety evaluations, red-teaming, safeguards and mitigations, failure modes, and evaluation methodology to identify where current approaches are strong and where important gaps remain." That is work judged by whether it controls an outcome, and the posting frames it that way itself.
A tension to govern,
not a ceremony to mock
The role will "Manage substantive technology-company advising work" and also "contribute to the development of new initiatives, partnerships, proposals, and funding opportunities," some of which may involve the same technology companies and philanthropies.
Advisor, evaluator and fundraising partner to the same industry is a workable position, and a common one in this field, but only when the independence rules are explicit. The public record does not show them. That is a governance item, and it is in the ledger below rather than in a verdict here.
Boilerplate that
does not control
The location block contradicts the posting's own designation. Small in itself, and precisely the pattern this audit exists to name: text that exists to reassure rather than to decide.
Engine two: the management-decision ledger
| The decision | Evidence | Belongs to |
|---|---|---|
| Who owns JED's internal AI adoption and its own data governance | The posting is entirely outward-facing; no public document names an inward owner | A named operations or technology owner, before staff adoption happens informally |
| The independence rules for advising companies JED may also partner with or be funded by | Advising and "funding opportunities" responsibilities in the same posting | Executive leadership, written before the first engagement |
| The recruitment and retention plan for the stated band | $120,000-$160,000 for seven-plus years of AI safety and evaluation expertise | SVP, Youth Strategies, with the board's compensation visibility |
| Which title and which scope the role actually carries | Two different names inside one posting | Settled with the first hire, in writing |
A row with no owner is not a decision, it is a hope. This table is meant to be signed.
Engine three: the people arc
The posting already names the skeleton: the Director collaborates with Knowledge & Advising, Policy, Programs and MarComms "to bring relevant expertise into the work."
The risk in every expert hire is the silo, one person who knows surrounded by teams who route around them. The method's answer is the same here as anywhere. Those named teams document what they do all day in their own words, see which parts AI already touches, and the Director's expertise lands inside those workflows rather than beside them.
For JED's own staff the inward question is not hypothetical. Programs, screening tools and communications will involve AI whether or not anyone governs it. Equipped, then oriented, then builder, measured the same way every month.
Engine four: the staffing map
Read against the five archetypes, this posting describes a Grower - field-building, partnerships, bringing priorities into shared frameworks - with a strong Sweeper component, since the evaluation and red-teaming mandate is gap-finding work.
What the posting does not describe, and no public JED document does, is a Builder for the inward half: someone who makes JED's own tools, governs JED's own data, and turns the Director's frameworks into JED's own practice. One hire cannot be both, and the posting does not pretend otherwise. The map simply shows that second seat, or the second half of a deliberate plan, as open.
The ninety days, outside-in
By day 30
The front door. jedfoundation.org does the basics well: the homepage carries full Organization structured data with the social profiles attached, and ten well-formed Open Graph tags. Credit where it is due, and it is more than most.
What is absent is any decision about answer engines. There is no llms.txt, no rule for AI crawlers in either direction, and a robots.txt of 177 bytes whose every directive is commented out, leaving a bare sitemap line in force.
For most organizations that is housekeeping. For JED it is mission-adjacent. Young people in distress increasingly put their first question to an AI assistant, and whether those answers draw on JED's guidance depends on exactly this plumbing. Making JED's resources legible to answer engines may be the highest-leverage safety work available for the cost of a config file.
By day 60
The inward charter. One page naming who owns JED's internal AI use and data governance, and the written independence rules for technology-company advising. Both are one-meeting decisions that become hard retroactively.
By day 90
The first decision-ledger review, and the role's settled scope published under one title.
The commitment this document honors
The Kurultai does not produce headcount-reduction recommendations, ever. This edition adds a second commitment in the same spirit: it treats the mission with the care the mission demands.
Every finding above is offered as a strategist's reading of a public record, to an organization doing work that matters, from someone who built the instrument because the work mattering is exactly why the strategy should be sound.
Sources
- Director, AI Strategy & Mental Health, full posting text from JED's own hiring page, retrieved 1 November 2026; application deadline 14 November 2026
- IRS Forms 990, The Jed Foundation, EIN 13-4131139, tax years 2011 to 2023, via ProPublica Nonprofit Explorer
- jedfoundation.org robots.txt and homepage markup, live inspection 1 November 2026
What this edition does not read. JED publishes its own positions and guidance on AI. This document refers to that work only as the posting itself describes it, and does not quote or assess it. A later edition would.
All public sources. No non-public information was used or sought. Every quotation above was checked against the posting as JED publishes it.