Why we onboard firms in cohorts (and sometimes say not yet)

A general counsel at a European industrial group signed with us in late August and asked to start the following Tuesday. Her team was ready, she said, and so was she. We told her the next cohort opened on 21 September. She was not pleased, and she said so, and three weeks later she started with eleven other legal teams and, by her own account at the end of the first month, was glad of the wait.

We onboard in cohorts. We have done so since early 2025, when the number of firms joining each month first exceeded what our implementation team could support one at a time. What began as a capacity constraint has become something we would keep even if capacity were unlimited. This is why.

What a cohort actually does

A cohort is eight to twelve legal teams starting in the same fortnight, following the same four-week sequence, with a shared weekly session in addition to each firm’s own implementation lead.

The sequence is the same for everyone. Week one is playbook capture: the firm’s lawyers sit with ours and encode their positions, fallbacks and red lines for the first practice area. Week two is calibration: Review runs against a sample of the firm’s own recent documents, the lawyers read every flag, and the playbook is corrected. Week three is live use on real matters with a narrow group of reviewers. Week four is widening to the full team and, usually, a second practice area.

The shared session is where the cohort earns its keep. Firms in the same cohort hit the same problems in the same week. In week one, nearly every firm discovers that a position it thought was a red line is actually a strong preference with three known exceptions. In week two, nearly every firm sees a flag it disagrees with, argues with it, and finds that the playbook said what the flag said. Hearing another firm describe the same experience shortens the argument considerably.

We have found that firms learn more from one another’s calibration mistakes than from our documentation, and that a cohort of peers removes a certain defensiveness from the process. Nobody wants to admit to a vendor that their playbook was ambiguous. Admitting it to a counterpart at another firm, who has just admitted the same thing, is easy.

Why not start on Tuesday

Three reasons, in order of how often they apply.

The first is partner time. Playbook capture requires senior lawyers who know the firm’s positions, not just the associates who apply them, and it requires them for several hours in week one and again in week two. Partners’ diaries fill. A firm that starts on an arbitrary Tuesday starts with whoever is free, and the playbook reflects that. A cohort date three weeks out gives the firm time to book the right people, and the shared schedule makes the booking easier to defend internally.

The second is the data we have on what happens otherwise. Firms that onboarded individually in 2024, before cohorts, took a median of seven weeks to reach the accept-rate plateau that cohort firms now reach in four. The difference is not the software, which was in some respects worse then; it is that the sequence was improvised each time, and the calibration week was frequently skipped by firms eager to go live.

The third is our own capacity to notice problems. Twelve firms following the same sequence surface patterns that one firm cannot. If four firms in a cohort stumble on the same playbook-editor screen in the same week, that is a product defect and we fix it. If one firm stumbles, it looks like a training issue.

When we say not yet

Occasionally we decline to onboard a firm into the next cohort at all. It happens perhaps once in every two cohorts, and it is never comfortable. The reasons are specific.

No playbook owner. If the firm cannot name a senior lawyer who will own the playbook, attend the capture sessions and make the calls in calibration, the onboarding will produce a playbook nobody stands behind, and the reviewers will learn to ignore it. We ask the firm to find the owner first.

The document management system is not ready. Review and Recall both depend on the DMS for permissions and matter structure. A firm mid-migration between systems, or one whose access control lists do not reflect its actual ethical walls, cannot safely connect. We would rather wait than index documents against a permission model that is about to change.

The firm expects autonomy. Some prospective clients describe what they want in terms that make clear they expect the system to review documents without a lawyer reading the output. Review proposes; a lawyer disposes. A firm that intends to skip the second step will have a bad experience and, more importantly, will be taking a professional risk we are not willing to help with. We say so, and we say not yet, and sometimes that becomes not at all.

The cost

Cohorts are slower for the individual firm. The general counsel who wanted to start on Tuesday had a real backlog and a real reason for urgency, and three weeks of waiting cost her something. We offer a short interim: Open Bar-style access to Recall against the firm’s own documents, with no playbook, so that search is available while the cohort date approaches. It is not the same as being live.

We also lose some prospective clients to the wait. A firm that is comparing options and is told “three weeks” by us and “tomorrow” by someone else will sometimes choose tomorrow. We have decided that is acceptable. A firm that goes live badly is a firm that goes live once.

The cohort calendar is published a quarter ahead. If the next date is inconvenient, the one after it is usually less so, and the playbook will be better for the delay.