How the move from seat pricing to usage pricing in legal AI affects law firms and in-house teams
What has happened
Two of the main AI companies, Anthropic and OpenAI, have started charging some customers by how much they use, rather than a flat monthly fee. Legora, a legal AI provider, has just done the same thing with its newest product. You pay for the work the AI does, not per seat.
Harvey, Legora’s main rival, still charges per seat as the default/standard today. But Harvey is still built on the same underlying AI models, and those models are not getting any cheaper to run. A move to usage-based pricing at Harvey looks likely, though not yet confirmed.
Both providers already sell a mixed version of this. You can buy a seat for unlimited use, or top it up with pay-as-you-go credits. So this is less of a sudden change and more of a shift that has been building quietly for some time.
Why this matters
Under the old model, your AI costs were fixed and easy to plan for. You knew what you were paying each month, whatever you used it for.
Under the new model, your bill moves with your usage. That sounds fair in principle. In practice, it means your costs become harder to predict, and the firms or teams that manage usage well will get more for their money than those that do not.
Will this favour the bigger firms?
To some extent, yes, but not simply because they have deeper pockets.
The firms that gain the most will be the ones that learn how their teams actually use AI, build some basic controls around that usage, and use their scale to negotiate better rates once they understand their own patterns. All of that takes time and resource. Bigger firms generally have more of both.
A smaller firm with disciplined habits could still get excellent value. But the broad effect is likely to widen the gap between firms that manage their AI spend well and firms that do not, and that split tends to track firm size.
Does this help legal tech companies bypass law firms altogether?
This is the part we think deserves more attention than it is currently getting.
A pricing model based on usage does not care who is doing the using. A law firm and an in-house legal team can both be metered the same way. That removes one of the few advantages firms have held over their clients, which is the ability to buy access in bulk.
We are already seeing legal tech providers build products aimed directly at in-house teams rather than at firms. The pricing shift makes that strategy easier to execute, not harder.
Does in-house buying power change the relationship with panel firms?
In some areas, yes. A large in-house legal team can often negotiate better usage rates than the panel firms it instructs, simply because of the volume it’s “client” company buys. If that team has generous internal access while its panel firms are watching every pound of AI spend, the in-house team can take on more first-pass work itself.
But this matters most for routine, high-volume work such as standard contract review or first-pass due diligence. It matters far less for complex matters that depend on judgement and risk advice, which is where firms, for now, still earn their keep. Having more AI access does not give an in-house team better legal judgement. It gives it more capacity to do the simple work in-house.
Does this make it harder for firms to keep pace with their clients on AI?
We think it does, and for a reason that goes beyond cost.
A fixed monthly fee is easy to budget for. A usage-based bill is much harder to forecast, because legal work varies enormously in size and complexity from one matter to the next. A firm that wants to roll AI out widely now takes on open-ended cost risk it did not carry before.
That makes broad rollout a harder decision to sign off internally. A firm either caps usage, which slows adoption, or accepts the uncertainty, which finance teams are unlikely to welcome. Clients may not face the same constraint if their AI spend sits inside a wider technology budget rather than a new line item under separate scrutiny.
A few other angles worth watching
Who absorbs the cost. If AI spend becomes variable and tied to specific matters, firms will need to decide whether to absorb it into their fees or pass it through to clients as a cost. Both choices come with friction.
Easier cost comparisons. As the underlying AI providers publish their own usage prices, it becomes easier for clients and firms alike to work out how much of a legal AI bill is genuine added value, and how much is markup.
A way in for smaller firms. Usage-based pricing could lower the barrier to entry for smaller firms currently priced out by minimum seat counts. That could widen the market at the smaller end, even as it widens the gap at the top.
A new internal job. Tracking and managing AI usage in real time is a new task. Someone, whether in finance, IT, or a dedicated role, will need to own it, in the same way firms already manage billable time and work in progress.
Our view
The direction of travel favours firms and teams that manage their AI usage with discipline, not simply those with the biggest budgets. Usage pricing turns AI from a fixed cost you can ignore into a variable cost you have to manage, and it rewards the firms that deploy it where it pays and scale what works, not those that simply buy the most seats or chase the highest usage. The risk of clients bypassing firms for routine work is real and worth planning for now, not later. In-house teams gain the most ground on commoditised work, not on the complex matters that still depend on legal judgement. And unpredictable costs may slow firms down at the exact moment their clients are moving faster.
Two practical points follow. Managing usage well does not mean using AI less: ration it too early and you starve the experimentation that shows where it genuinely pays, so budget deliberately for that learning. And while metering hands you a cost dashboard for free, the harder and more valuable discipline is measuring what the usage is worth, not just what it costs.
We would be glad to talk through what this means for your own AI strategy and contract terms with your providers.