Case study
I founded the Illinois Data Equity Project. Its governance is being built so that founder-control gives way to community leadership over time, by design. The interim has commitments. The hand-off has triggers. This is the methodology applied to its own institution.
What IDEP is
A 501(c)(3) that turns public policing data into analysis tools, educational modules, and organizing pathways for the communities most affected by policing decisions. The platform is built on data-sovereignty and community-empowerment principles, with a graduated consent system, and operationalizes the CARE principles for Indigenous data governance (Carroll et al., 2020) in its consent and access decisions.
The case study below is not about the platform. It is about how the organization that holds the platform is governed — and why its founding governance is designed to fold itself into community leadership.
The premise
Community-led is not possible at founding, when a person or small group builds something they find useful. The community didn't ask for it. You did.
Every other claim about community governance falls out of that one. Most non-profits paper over it with a board and a mission statement and call the question closed. The honest version is messier — and more durable, if you build for the mess from the start.
The design
Piece 1
The founding team's decision authority is not designed to extend indefinitely. The transition is something we are instrumenting in the governance — not deferring to "when it feels right" or letting drift onto a wiki page that rots.
Piece 2
The transition to stakeholder-led governance is keyed to named conditions: stakeholder body convened, stakeholder authority exercised in practice, scope and budget understood by people who are not me. The work is to make each condition observable and dated, rather than leave the call to vibe.
Piece 3
Full transition isn't possible on day one. So the design specifies what should be true on day one anyway: scope locked to the founding public-interest charter, decisions recorded with their reasoning preserved, stakeholder input pipelines built before they are needed, transparent reporting on what the founder-led period is actually doing with founder-led authority.
What this is not
It is not a pledge to listen better. It is not a stated intention to be community-led "eventually." Pledges and intentions are the failure mode this design is built to replace. The bet is structural: that the conditions of the transition can be written down at the founding moment, not discovered after.
That is what should make the design durable across the founder leaving, stepping back, or being replaced — and legible to the stakeholders who are eventually meant to govern it. The work is in instrumenting the bet, not just taking it.
Why this matters for AI governance
The person who writes today's AI policy is not the population the policy will govern tomorrow. The team that designs a model's review process is not the workers, customers, or patients whose lives the model will reshape. The asymmetry is built into the work.
Most AI governance pretends the asymmetry doesn't exist. The framework gets written. The audit happens. Affected populations are, at best, consulted. When the model updates, when the regulator changes posture, when the team that wrote the policy moves on — the people downstream of all of it inherit the consequences. They were never structurally part of the governance.
Governance shelf-life takes the asymmetry as the founding constraint, not an awkward footnote. The IDEP design is one example of what that looks like in practice. The diagnostic is the same lens, applied to your governance.
Want to see how this lens applies to your governance?