How Do We Run Cohorts to Build and Scale Tech Solutions?
This is a guide to running cohorts in this space, and to the distinction that decides whether a cohort works: whether it is a Generator that produces strategic clarity, or a Carrier that moves working tools between organizations. It explains both roles, why naming the role up front matters, and how to design and fund for each.
Common Legal Help AI. Working draft for comment. Margaret Hagan, Stanford Legal Design Lab.
The cohort that produced this body of work was a particular kind of cohort, and naming the kind matters, because the next ones should be designed on purpose rather than by default. The clearest lesson about cohorts is that they do two different jobs, and a cohort that tries to do both at once tends to do neither well. This guide is that distinction and what follows from it.
Two kinds of cohort
A Generator cohort produces ideas, strategic clarity, frameworks, and named needs. Its outputs are conceptual. It runs over a longer arc, it expects looser shipping, it builds trust through sustained discussion, and it is the right shape when a field does not yet know what to build. The TIG cohort was a Generator, and it produced what Generators produce, which is the strategic frame the field now operates with.
A Carrier cohort moves an existing working practice, tool, or pattern from one organization to others. Its outputs are operational. It runs over a shorter arc, it expects regular shipping, it draws on trust already built, and it works through paired engineering and active externalization. It is the right shape once working examples exist and the field's question has moved from what to build to how to spread what works.
Why naming the role up front matters
The role a cohort plays decides its time horizon, its meeting cadence, the capacity each site needs, its shipping pace, and what counts as an output. Funders and participants should agree on the role before recruitment, because the recruitment criteria differ. A Generator wants diverse, interested participants who can sustain attention over a long arc. A Carrier wants attorney-and-engineer pairs at each site who can ship working tools quickly. Mixed recruitment produces uneven outcomes, which is the failure mode of a cohort that never decided what it was.
Designing a Carrier cohort
A Carrier cohort needs things a Generator does not. It needs engineering capacity paired with facilitation, because a single facilitator from a convening organization can host a good conversation and cannot build and ship a shared tool. It needs externalization built into the operating model from day one, with a named publication function whose job runs continuously, so that every meeting produces something the field can use instead of something that stays in the meeting room. And it needs its carrier portfolio identified at the start, meaning where the outputs will be published, who will adopt them, what recognition flows back to participants, and what funds the next cycle.
The two-stage pattern
The cleanest version of this work runs in sequence. A Generator cohort produces the strategic clarity and identifies what needs to be built, and then Carrier cohorts build and spread the specific tools, drawing on what the Generator produced. The funding can follow the same shape: one grant to surface the question and the frame, then a follow-on grant to ship the working tools. The TIG program funded the Generator here, and future grants could fund the Carriers. The two roles work better in sequence than they do collapsed into one cohort that tries to carry both at once.
Where each kind fits now
Both models have a place. The Generator model suits the early stage of any new technical area, when the field does not yet know what to build, which is what AI for legal aid was when this cohort started. As new technical questions arrive in the coming years, Generator cohorts of similar shape will likely be the right tool for them. The Carrier model is what this particular field needs next, because working examples now exist across multiple state deployments, and the open question has moved from what to build to how to spread it. That is the cohort the field should design and fund next, and the field guide is where the case for that shared, ongoing work sits.
The most durable output
The most durable thing a cohort produces is often its people. The advisor and practitioner network a cohort builds can outlast the grant, with participants serving as advisors, mentors, or paid contributors to the cohorts that follow. Treating that network as a continuing output rather than a one-time input is one of the surest ways a cohort's value keeps traveling after the grant closes. The Working Groups are where that continuing network lives.