Per-seat pricing is broken for legal AI, and here is the math
Why charging per lawyer makes firms hoard licences and kills adoption, and what a credit model has to get right to be better rather than merely different.
Per-seat pricing works when usage per person is roughly flat. Legal AI is the opposite: a firm has a handful of people running reviews constantly and a long tail who need the tool four times a year. Price it per head and the finance conversation becomes a licence-rationing exercise.
What rationing does to adoption
The paralegal who would have caught the missing indemnity does not have a seat. The commercial head who wants a second read on a supplier agreement does not have a seat. Both route around the tool. One does the work manually; the other pastes a clause into whatever consumer chatbot is open. The firm now has a lower adoption number and an uncontrolled data path.
That second outcome is the expensive one. Seat scarcity does not reduce AI usage in a firm. It relocates it somewhere with no audit trail.
The math a credit model has to get right
Charging for work rather than for headcount only helps if the unit price tracks the actual cost of the work. In LexCore a credit is ₦150 and the dominant cost driver is the review mode, not document length: a single grounded pass is cheap, a full adversarial run across three providers is not. Each debate round after the first costs about 110% of the first, because every round re-sends the accumulated argument. Cost grows faster than round count, and rounds are capped at five regardless of plan.
Seats do not disappear, they stop being the meter
Plans still carry seats, because access control is real and a firm needs to know who can open a matter. What changes is that a seat is no longer the thing being rationed: the allowance is credits, and the allowance is sized to cover a working month rather than to run out in week three. The full catalogue is on the pricing page: plans, credit costs per mode and size band, and an estimator.
The part vendors skip
When a trial ends without a subscription, the account stays open and free, indefinitely. Matters, documents, past reviews and the audit trail remain readable; only running a new review needs credits. Locking a lawyer out of their own matter history to force a purchase ends the relationship permanently, and it costs nothing but storage to not do it.
Keep reading
The sovereignty tier: running local inference for banking and government briefs
Hardware, latency and the tradeoffs of running the whole stack inside your own network, including the one requirement people try to skip.
Building a house playbook when your partners disagree on everything
A method for turning a shelf of closed matters into one tiered position document a partnership can actually live with, and what to do about the clauses nobody agrees on.