Legora's move from flat, unlimited pricing to token-based consumption marks the first crack in legaltech's pricing model. AI inference costs are real and variable in ways seat licensing never was, shifting the economics for everyone in the chain. For two years, the story has been "drive adoption at any cost." Consumption pricing inverts it. Once every query runs a meter, maximum usage stops being an unambiguous good: buyers want to control it, vendors have to fund it, and go-to-market teams now have to justify it rather than celebrate it. That is the adoption paradox, and it reaches all six groups below.
Just because a company makes lots of money or has raised lots of money doesn't mean it is worth lots of money. With Legal AI, the more people use it, the more it can cost the company to run.
Usage-based revenue forces investors to weigh cost-to-serve — the inference bill — alongside revenue, and gross-margin volatility becomes a live diligence question.
We heard at a recent Law Department Operations Survey roundtable that some law departments are already trying to get ahead of AI cost increases by experimenting with open-source models.
Consumption pricing reopens the valuation question. Does legaltech now trade like an infrastructure business — lower multiples, harder unit economics scrutiny — rather than a SaaS company priced on ARR?
The AI companies that win will be the ones that can grow quickly and make good money from that growth. If more people use it and you just spend more to run it — that's not a business.
Usage-based pricing demands forecasting, metering, and customer education that most legaltech companies have never built. Get it wrong and you either scare buyers off or underprice the work.
Executives can't allow unfettered spending on AI any more than they can allow endless spending on kitchen snacks. Promises of efficiency that result in a negative impact on EBITDA are unsustainable. Access could concentrate.
Usage-based pricing complicates the fundraising story: "unlimited usage drives adoption drives expansion" was a clean line, and consumption pricing ties revenue to usage that no one can predict. But the same frustration works in a challenger's favor.
If the foundation models turn expensive and unpredictable, simple, capped, or flat pricing becomes a real differentiator. The open question is whether consumption pricing works for point solutions at all — a narrow tool may never reach the daily volume a broad platform uses to absorb variable costs.
A few visible surprise bills could flip buyer behavior from deployment to justification — and cycles will lengthen as procurement and finance enter the room earlier.
Overpromising ROI or underestimating real costs during a sales cycle will become more detrimental as time goes on. With a new technology like AI, it becomes existential.
Firms and departments have raced to adopt AI with cost as an afterthought. A few visible surprise bills could flip that behavior from deployment to justification. Two forces compound the shift.
If you tell a customer AI will save them £1 million, but the bill ends up being £1.5 million, they won't trust you again. In AI, trust is becoming one of the biggest things you can sell.
According to Bryan Catanzaro, Nvidia's VP of applied deep learning, AI compute costs for his team now far exceed what the company spends on the employees using it.
The low-friction pilot may become ineffective if buyers believe costs will increase after the trial. Marketing will require new evidence: cost guarantees, usage caps, or a reliable pricing calculator.
As the market moves past a "use AI for whatever we can" philosophy to one with guardrails, marketers will have to focus less on adoption and more on effective usage, ability to embed in workflows, and ROI.
Marketing can't just say, "Look how much AI our product uses." Buyers will want to know, "What do I get for my money?" Show me what it saves, show me what it makes, and show me what it will cost.
It's also strange that the big LLM companies haven't created a simple, universal way to explain how powerful their models are. We understand horsepower in a car. But with AI, we're given confusing model names and scores that mean very little to most buyers. If you can explain the value simply, people will trust you.
Consumption pricing rewrites the most prized go-to-market and customer-success profile. For two years, the valued hire was whoever could drive usage up. Under a meter, that instinct becomes a liability.
This is why so many companies are suddenly interested in Legal Engineers. The job is changing. We used to sell a five-year contract and hoped the customer would stay. Now, the work starts on day one. Legal Engineers help customers get real value from the technology, control costs, and make sure they don't leave. The future isn't just about selling the contract. It's about protecting the contract from churning.
Customer success and post-sale roles shift from cheerleading adoption to defending ROI and managing spend: a different skill set entirely. More usage can now lead to bill shock and churn — the instinct to maximize usage must be replaced by the discipline to maximize value.
Token growth turns AI delivery into a variable, rising cost of goods rather than a fixed one — the math that pushes vendors toward consumption pricing and puts headcount under pressure.
The underlying assumption that work done with AI is less expensive than work done by humans may end up challenged as AI costs continue to rise.
Legaltech startup Darrow cut roughly a third of its staff — about 60 roles — though the company says it stayed profitable and called the move a reorganization.
Many of those cut were the legal analysts whose work trained its AI: a reminder that a maturing model can displace the people who built it. It remains unclear if startups will consistently trim while full-stack legal AI grows.
The teams that win will be the ones that bring lawyers, engineers, and salespeople together. More lawyers are helping with sales, more legal counsels are getting involved in marketing. The old sales team is changing.
Each quarter, we translate market data and insight into actionable signals for legaltech leaders.