AI Legal Software Flails on Pricing Models, Frustrating Buyers

Legal technology’s already erratic pricing landscape is on the verge of more change as consumption-based models threaten to confound in-house legal teams trying to figure out what to buy.

Traditional software pricing, where vendors charge on a per-seat basis, remains dominant. But as AI becomes more integral to software, a few companies are starting to charge based on how much a person uses AI tools or how many documents a user plans to process via the software.

The new pricing models could make it harder for tech buyers to compare how much one tool will ultimately cost versus another. Companies are already struggling to demonstrate a return on investment on their AI tools, as vendors have difficulty putting a dollar value on their products.

“The story is, ‘Buyer Beware’,” said Lucy Bassli, a consultant to in-house legal teams. “It is completely all over the place. It’s changing within the same vendor. You can do a demo, start negotiating, and it’ll change during that. You can have a peer that’s going to get a completely different deal because it was one month they signed earlier or one month later. It is changing that quickly.”

“It’s hard enough to pick the right solution from a capabilities perspective. Once you get into pricing, it is all over the place,” she added.

AI’s mega companies, Anthropic and OpenAI, have shifted their pricing models over the years. The thousands of smaller companies building tools around their large language models, Claude and GPT, have to adapt to those changes while figuring out themselves how to make money.

“We’re all flailing, everyone’s trying to figure out how to do any of this now,” said Nicole Bradick, global head of innovation at the legal services company Factor.

Consumption Pricing

The big AI companies measure usage via “tokens.” In legal tech, a consumption-based pricing strategy accounts for how many tokens a lawyer uses. It’s unpredictable, meaning software costs could swing month to month. The advantage is lawyers only pay for what they use.

The idea is new enough that it hasn’t been widely tested among cautious attorneys, who might be less willing to experiment with AI because they could run out of tokens or incur additional charges, said Patty Corey, legal operations manager at Franklin Templeton.

Innovation leaders tell lawyers to think about AI as a “thought partner,” but pricing per “thought” undermines that strategy, she said. Attorneys who don’t know how to use AI efficiently could run out of tokens under their license or face higher charges, she said.

“You’re telling me to experiment with AI, and if I don’t do it properly, then I’m going to be penalized?” she said.

One way to answer that concern is to offer tiered subscriptions that are relatively easy to move between. Each subscription level would provide different amounts of AI computing, said Otto Hanson, vice president of product at the contract management company Agiloft.

If users need more computing power, they can easily get it. And if they find they use less, they don’t need to pay for more than what’s necessary.

“You’re really only paying for what you use and that is, I think, a very fundamentally customer friendly thing to do,” Hanson said.

The contract technology startup Ivo is exploring a similar model. The company prices for contract review on a per seat and “occasionally but increasingly on a consumption basis,” CEO Min-Kyu Jung said in an email. For its contract intelligence platform, Ivo charges based on the number of contracts a customer puts into the platform.

“That’s the unit of value: a contract the business can query, analyze, and act on,” Jung said. “Pricing scales with the size of that portfolio.”

Seat-Based Concerns

A seat-based model, still the standard in the industry, offers better pricing consistency but can also dissuade adoption. When companies are paying per seat, they tend to limit the number of licenses they buy, said Stephanie Corey, the former head of legal ops at Hewlett-Packard.

“When you’re looking at a seat license for lawyers who live in all of those tools, it can really kind of dissuade teams from getting broader adoption across legal because that could get very expensive,” Corey said.

That’s not ideal for vendors either, who want to get as many people as possible hooked on their platforms, Hanson said.

Like Ivo, Agiloft has different pricing models depending on the product. Agiloft offers pricing by the seat, but its AI contracts platform Astra is experimenting with consumption-based pricing.

“We’re trying to figure out the best way to price and package our services in this new world,” Hanson said.

Impact on Vendors, Departments

Both buyers and sellers of the tech are grappling with uncertainty. The AI hype wave hasn’t required tech builders to make money yet, but that’s not going to last forever, and AI companies will eventually need to account for the high computing costs tied to their products.

“Token consumption is going to be a huge problem if you’re not pricing for that,” Bradick said. “If you don’t care about being profitable, it’s not a problem.”

On the buyer side, companies can’t show a return on investment on their AI purchases, Stephanie Corey said.

“I don’t know what the right model is yet, because we haven’t kind of come together to decide that,” she said. “Because of all of this craziness right now, it is hard to prove an ROI.”

Buyers are annoyed, but that’s par for the course with new technology, Bassli said. Widespread generative AI is just a few years old, and many of the legal tech companies selling to in-house teams are offering products that are only months old.

“There really is just frustration in the market, and it’s not to blame the tech,” she said. “The tech’s moving at lightning speed. So it is what it is right now.”

Bloomberg Law is a legal technology provider.