Let's build the public-goods layer together
Independent evidence about AI and work should belong to everyone. We're building a free, open layer — a workforce index, an open assessment framework, open-source tools, and research on keeping human oversight competent — and we're looking for nonprofits, academic centers, and foundations to build it with us.
Open by default · Methodology-transparent · Non-exclusive · No vendor capture
The evidence about AI and work shouldn't be owned by the people selling AI
AI is reshaping work faster than anyone can independently measure it. The organizations best placed to document that shift honestly — universities, nonprofits, foundations, and civil society — are rarely the ones with the assessment infrastructure to do it at scale. Meanwhile the most detailed data sits with vendors, who have every incentive to tell a flattering story.
We'd rather build the alternative in the open, with partners, than sell it. Everything in this layer is free to use, transparent in method, and open to challenge.
Looking for something else?
If your organization wants to build and run its own credential, that's a different relationship — see Evidentia. This page is about co-building the free, shared layer.
What we're building — and want help building
Each of these is free, openly licensed, and stronger with more hands on it.
AI Oversight Workforce Index
A free, annual, methodology-transparent picture of which AI oversight and assurance skills are growing or shrinking, role by role — and where the gaps are.
Open assessment framework
A public, versioned, citable framework for how AI oversight competence should be defined and tested — free for any body to adopt or adapt.
Open-source readiness tool
A self-assessment that maps a person or team against oversight competencies, surfaces gaps, and points to ways to close them.
Evidence Briefs
Plain-language, role-by-role explainers on how AI is changing specific jobs and what meaningful human oversight looks like in each.
Anti-deskilling research
Studying how automation erodes the practiced judgment oversight depends on — including the question no one can answer alone: if AI removes the entry-level rungs where judgment forms, where do the senior overseers of the next decade come from?
Public credential register
Open verification infrastructure so any credential built on the framework is publicly checkable.
Public-interest advisory
Guidance for nonprofits, policy bodies, and coalitions who need independent counsel on AI oversight but can't fund commercial consulting.
Open reskilling curricula
Freely licensed curricula for moving workers whose tasks are being automated into AI oversight, evaluation, and assurance roles.
Five kinds of partner
Nonprofits & civil society
Organizations working on workforce, labor, equity, education, or digital rights. You bring reach, lived understanding of who's affected, and a check on whether our work actually serves them.
Together: co-designed programs, joint distribution, shared advocacy.
Academic centers & research labs
Groups studying AI's labor-market effects, human-AI interaction, automation bias, or assessment science. You bring methodological rigor and independent peer scrutiny.
Together: co-authored research, methodology review, shared datasets, student projects.
Foundations & funders
Funders backing future-of-work, AI safety and governance, or economic mobility. You make it possible to keep this layer free for the people who need it most.
Together: fund the index, the open tools, and reskilling pilots — with findings independent of funding.
For-profit companies & AI developers
Companies deploying AI at scale, and the developers building it. You bring deployment reality, evaluation evidence, lab environments, and demand for credentialed oversight staff — all things the commons can't generate on its own.
Together: contribute evidence, infrastructure, and demand — under published terms that keep findings independent. See below.
Career & workforce platforms
Platforms that read capability and route people toward new work. You bring gap detection, demand data, and reach into people mid-transition — a pathway is only real if it ends somewhere verifiable.
Together: open competence frameworks and publicly checkable credentials your routing can point to — verified destinations, not just directions.
We're also glad to hear from professional societies and standards bodies and workforce boards.
Contribution without control
Companies hold data, infrastructure, and hiring demand the commons genuinely needs. The question is never whether to work with them — it's on what terms. Here are ours, in public.
1 · Lab & sandbox environments
Supply the environments practical assessments actually run in — model access, tooling, compute. You gain a trained operator workforce and a hiring signal; the commons gains infrastructure at near-zero cost.
Minimum three providers per capability class, rotated annually.
2 · Evidence contribution
Publish your evaluations into the commons and let them be independently appraised. Self-published evidence, however rigorous, isn't independent assurance — independent appraisal is exactly what makes it credible. You get standing you can't generate alone.
3 · Demand-side commitments
Pilot reskilling cohorts, certify your own AI-oversight staff against the framework, and hire credential holders. This is the highest-leverage contribution any company can make — demand is what gives a credential weight.
4 · Funding the commons
Support labs, scholarships, and convenings — downstream of methodology, never upstream of it. Corporate capital never reaches curriculum, competency definitions, or exam design.
The guardrails, published
- No single-vendor dependency. At least three providers per capability class; environments rotate annually.
- No endorsement. Participation buys no preferential placement, ranking, or implied approval.
- Open primitives only. Competencies are defined against open standards, never a single vendor's tooling — ours included.
- Firewall on methodology. Corporate support never touches curriculum, competency definitions, or exam design.
- Full disclosure. Every corporate contribution is publicly listed.
- Editorial control stays with us and our independent partners. Findings are findings.
Five ways in
| Engagement | What it looks like | Best fit |
|---|---|---|
| Research collaboration | Co-design and co-author studies; peer-review our methodology before we publish | Academic centers |
| Co-build a public good | Jointly develop the index, framework, briefs, or open-source tooling | Nonprofits, academic centers |
| Fund the commons | Underwrite a specific public good so it stays free and independent. ForEvidence.ai is a public benefit corporation, so grants are routed through a nonprofit partner or fiscal sponsor — we'll set that up with you | Foundations & funders |
| Contribute infrastructure, evidence or demand | Supply lab environments, publish evaluations for independent appraisal, or commit to reskilling and hiring — under the published guardrails above | Companies & AI developers |
| Distribution & pilots | Bring the tools to your community; run a reskilling or oversight-training cohort | Nonprofits, employers, workforce boards |
| Advisory & governance | Sit on the advisory group that reviews methodology and guards independence | All partner types |
The terms we hold ourselves to
- Open by default. Outputs are openly licensed and free to use, adapt, and republish.
- Methodology in public. We publish how we did it, including limitations, so anyone can challenge it.
- Funding never shapes findings. Funders support the work; they don't review or influence results. We say so in writing.
- Non-exclusive. A commons, not a lock-in. Partners can work with anyone else, on anything.
- Shared credit. Co-authorship and attribution for the people who do the work.
- No vendor capture. Nothing is built to steer demand toward a single vendor's tooling — ours included.
- Peers, not referees. We never evaluate, rate, or rank fellow public-good builders — a commons where contributors grade each other stops being a commons.
What we bring
- Assessment engineering — evaluation harnesses, practical exams, scoring design
- Evidence methodology and auditable, reproducible records
- Operating capacity to actually run programs at scale
- Platform & verification infrastructure to host the commons
What we're asking for: rigor, reach, funding, and honest scrutiny.
Tell us what you're working on
No formal proposal needed. A few lines about your organization and what you care about is plenty — we'll take it from there.
Or just email us
Reach us directly at partnerships@forevidence.ai.
For funders
We're happy to share a detailed brief on any single public good — scope, method, cost, and how independence is protected. ForEvidence.ai is a public benefit corporation; grant funding is routed through a nonprofit partner or fiscal sponsor so your support stays tax-deductible.
Building a credential instead?
If you want to stand up your own certification program, see Evidentia.
The commons is better with you in it
If any of this is close to what your organization already cares about, let's talk.