Is per-seat or usage-based pricing better for legal AI?
Per-seat vs usage-based pricing for legal AI comes down to who carries the cost of heavy use. Per-seat pricing gives a law firm a fixed, predictable bill for each lawyer; usage-based pricing charges for what is actually consumed, which can suit light users but is harder to budget. For transactional teams doing repeatable drafting and negotiation work, a published per-seat price scoped to that work is usually the easier model to plan around.
Legal AI is now a line in most firms' technology budgets, and the pricing model matters almost as much as the headline price. A tool that looks cheap per user can cost more once usage caps bite; a tool billed by consumption can look reasonable until a large matter lands. This post sets out how the two models work, where vendors such as Harvey, Legora, Spellbook, Luminance, Robin AI, Definely and Microsoft Copilot sit in the conversation, and the questions a mid-sized firm in the UK, Ireland, the USA or Canada should ask before signing.
What is the difference between per-seat and usage-based pricing for legal AI?
Per-seat pricing charges a fixed amount for each named user, usually per month or per year. Usage-based pricing charges for consumption, such as documents processed, queries run or the underlying AI tokens used.
Many vendors combine the two: a seat price with a usage allowance, and a charge or a slowdown once the allowance is used. The table below sets out the practical differences.
| Question | Per-seat pricing | Usage-based pricing |
|---|---|---|
| How you are charged | A fixed fee per user per month or year | By documents, queries, pages or tokens consumed |
| Budget predictability | High: the bill moves only when headcount does | Lower: the bill moves with matter volume |
| Occasional users | Can be poor value if a seat is rarely used | Usually good value |
| Heavy users | Good value, unless the contract caps usage | Can become expensive quickly |
| Who carries the risk of heavy use | The vendor, spread across all seats | The firm, matter by matter |
| Recovering the cost from clients | Treated as an overhead | Easier to attribute to a matter, if the client agrees |
A general comparison of pricing models, not a description of any one vendor's terms.
Why does usage matter so much in legal AI?
Because the underlying models are paid for by volume. Model providers such as Anthropic publish their API prices per million tokens, so every long document a tool reads and every answer it writes has a real cost to the vendor.
Legal work is unusually heavy on input. Reading a full data room, a disclosure bundle or a long research question uses far more tokens than drafting a standard clause from an agreed precedent. A vendor has to recover that cost somewhere: in the seat price, in a usage cap, or in a consumption charge.
Do transactional lawyers subsidise litigation users on broad platforms?
That is Scriptum's view, and we think it is a fair question to put to any vendor. When one flat seat price covers every practice area, it has to be set high enough to cover the heaviest users.
Broad platforms such as Harvey and Legora are built to serve research, review and drafting across a whole firm, including disputes work, and that breadth is valuable to firms that need it. But litigation and investigations work can involve very large document sets, while much transactional drafting starts from a firm's own precedents and is far more repeatable. If a corporate or property team pays the same seat price as a disputes team that runs the tool over thousands of pages, the transactional lawyers are, in effect, helping to pay for usage they do not generate. We made the wider argument in should a transactional firm pay for an all-purpose legal AI platform?
A flat seat price is only fair if the work behind each seat looks roughly the same.
Which pricing model suits a mid-sized law firm?
For most mid-sized firms, a per-seat price for a tool scoped to the work the team actually does is the most predictable option. Usage-based pricing can make sense for a small group of occasional users or a single large project.
Whichever model you are offered, work through these questions before you sign:
- Is the price published? If not, ask for a written quote on the same basis as other vendors, so you can compare like with like.
- What does a seat include? Check which features, agents and integrations come with each plan, and which are extra.
- Are there usage caps? Ask about limits on documents, pages, queries or tokens, and whether they are per user or pooled.
- What happens at the cap? Find out whether you are charged overage, slowed down or cut off.
- What is the minimum commitment? Check the minimum number of seats and the contract term, and whether you can start with one practice group.
- Can you mix seat types? A firm rarely needs every lawyer on the top plan.
- How are prices quoted? Confirm the currency, whether VAT or sales tax is included and how often you are billed.
Our guide to how much legal AI costs a UK law firm in 2026 covers typical budgets in more detail.
How do Harvey, Legora, Spellbook and other tools price?
Check each vendor directly, because pricing changes often and several vendors quote on request rather than publishing list prices. That is a legitimate commercial choice, but it means you will need quotes to compare.
Broadly, firms will meet three kinds of offer: broad legal AI platforms such as Harvey and Legora; contract-focused tools such as Spellbook, Luminance, Robin AI and Definely; and general assistants such as Microsoft Copilot, licensed through Microsoft 365. Ask each the same questions from the checklist above, and get usage terms in writing.
How does Scriptum AI price?
Scriptum AI publishes a per-user monthly price. At the time of writing, Starter is £79 and Pro is £119 per user per month, excluding VAT and billed monthly, with custom pricing for the Firm plan; the Scriptum pricing page also shows prices in euros, US dollars and Canadian dollars.
Every plan includes the Scriptum add-in for Microsoft Word, the matter workspace and a clause library that belongs to your firm. Drafting starts from your firm's own precedents rather than generic wording, and the knowledge captured as you work stays with the firm. Scriptum is built for transactional work at mid-sized firms with between 5 and 500 fee earners in the UK, Ireland, the USA and Canada, so the price reflects that work rather than a firm-wide mix of practice areas. The Draft product page shows how drafting works in Word.
Comparing legal AI quotes for a transactional team? Book a demo and we will walk through a live matter and the plan that fits.
Sources
- Anthropic: Pricing (API prices per million tokens)
- Scriptum AI: Pricing (checked 20 November 2026)
Common questions
Is per-seat or usage-based pricing cheaper for a law firm?
It depends on how evenly your lawyers use the tool. Usage-based pricing can be cheaper for light or occasional users, while per-seat pricing is usually easier to budget and avoids surprise bills when a busy month drives usage up.
Why do some legal AI vendors not publish their prices?
Many vendors price by quote because costs vary with the number of seats, practice areas, usage and contract length. That is a commercial choice, not a sign of a bad product, but it makes like-for-like comparison harder, so ask for a written quote on the same basis from each vendor.
What should a law firm ask about legal AI usage limits?
Ask whether there is a cap on documents, pages, queries or tokens, what happens when you reach it, how overage is charged, and whether limits are pooled across the firm or set per user. Get the answers in the contract, not only in a sales deck.
How much does Scriptum AI cost?
Scriptum AI publishes per-user monthly pricing on its pricing page. At the time of writing, Starter is £79 and Pro is £119 per user per month, excluding VAT and billed monthly, with custom pricing for the Firm plan. Prices are also shown in euros, US dollars and Canadian dollars.