The founder of one of legal tech's buzziest startups says a move away from the software seat was only "a matter of time."
Since the rise of cloud software,** **companies have mostly charged for it by the user. A business bought a subscription for a set number of licenses, or "seats," and paid roughly the same whether employees used it religiously or hardly at all.
Max Junestrand started Legora, which sells software to the legal industry, at the tail end of software's great subscription boom. Three years later, that playbook is starting to break down. Artificial intelligence is scrambling the economics of software. Legora has changed its pricing model in response, and other software companies are doing the same.
Every ask to an AI model uses computing power. Some requests take far more juice than others. Those swings in usage create a problem for the companies racing to bring these systems into the workplace: The more customers use them, the more expensive they can be to operate.
Legora is feeling that tension. In June, it released Legora Agent, a system designed to handle some of the work of junior lawyers from start to finish. Suddenly, one lawyer could rack up far higher computing costs than another, though both occupied a single seat. A spokesperson said the share of users who use Legora daily is 13% higher among Agent users than among those who don't use the feature.
Junestrand said it no longer made sense for a light user to pay the same as someone who "really gets the most juice out of all the Legora functionality."
Later that month, Legora shifted its pricing model. Existing customers would keep their current contracts for the core product, with the option to upgrade to its pro-tier AI agents on a usage, or "consumption," basis. New customers, however, would be put on pay-as-you-go pricing.
The response from law firms was "good" but "mixed," Junestrand said. Some customers welcomed the change because they could now link their software spending to specific matters. Junestrand gave an example: An in-house lawyer for a company can see what it costs to review a batch of documents or perform an analysis with Legora, then compare that with the cost of sending the same work to outside counsel.
Legora has also built a dashboard that lets customers track their usage and a calculator that helps them forecast their spending.
Other customers were more skeptical, especially those still figuring out how AI fits into their work. Junestrand said those firms are still testing Legora's standard AI agents, which come with their existing seat-based contracts, before deciding whether to move up tiers.
The shift is about more than how Legora charges customers. If customers use an AI product far more heavily than expected, the company that makes it can end up owing its model providers more and more money, without bringing in any additional revenue. For some startups, usage-based pricing may be less an experiment than a condition for survival.
Legora is far from the first software company to kick the seat. Coding startups Cursor and Lovable have moved toward usage-based models that charge power users more. The model providers underneath many of these products, including OpenAI and Anthropic, have charged developers based on usage from the start.
Legora's biggest competition, Harvey, isn't ready to give up the seat. A spokesperson said the company plans to keep offering seat-based pricing because it gives customers "transparency and predictability," even as Harvey explores other ways to charge customers who want to pay for additional usage or value.
Whether the seat survives may ultimately come down to which kind of uncertainty customers would rather live with: paying for software they don't use, or not knowing exactly what it will cost when they do.