In a partner article published by Les Echos, Pappers says its legal AI product, Pappers Justice, has grown 120% a year since it launched its paid activity three years ago and is winning ever-larger law firms. While the legal AI market filled with what it calls "ChatGPTs for law," Pappers made two different bets: build its AI on its own databases, and open those databases to clients so they can connect them to their own tools.
From search engine to production tool
Pappers Justice opened in 2021, following the model that has made Pappers a success since 2020 with company data: a free platform, here for court decisions. The turn came two years later with the arrival of Gabriel Moneyron, who was tasked with building a complete solution for lawyers and legal departments and setting up a subsidiary around it. Moneyron met Pappers cofounder Romain Banchetti at Centrale and had already launched a parliamentary open data project, now Pappers Politique.
It began as a search engine over a document base: court decisions indexed and cross-referenced with the parties, statutes, collective agreements, the BOFiP, the BOSS, international tax treaties and more. The team then added the daily work of a firm, from case synchronization to legal monitoring and automatic exhibit stamping with an index. The AI, which arrived two years ago, changed the nature of the product: the professional no longer searches for a decision but hands over a substantive question and gets back a sourced analysis built on the documents in their own case file. Research that took half a day can now take minutes, so the product moved from a reference database to a production tool.
Bet one: data before the model
Moneyron says "many legal products are only a thin layer of prompt engineering on top of a general model, resting on artisanal scraping of Légifrance and public decisions. That makes very good demos, but it hallucinates on a real case." Pappers has built its databases since 2020, combining open data, private sources and proprietary data. Beyond decisions and statutes, they include company, property and land-registry data and parliamentary data, and firms can upload their own cases and query them on a sovereign cloud. The hard part, Moneyron says, is linking a raw decision to the parties, their companies, the texts it applies and the decisions before it. "Our AI does not invent a decision, it goes and gets it": it chains searches, verifies and runs complete workflows, helped by the depth of the Pappers Immobilier, Entreprise and Politique bases.
The first subscribers were small: solo lawyers, firms under ten people and accountants buying a licence for themselves. The launch of Pappers IA accelerated things, since a tool that can run a complete search interests teams. According to Moneyron, the larger the firms Pappers meets, the more its legal AI convinces them: they are the most demanding on sources, hosting and what happens when the model does not know, and they now account for most of its revenue.
Bet two: put the data in clients' tools
Legal data is traditionally a fortress, with each publisher protecting its collection for its own interface. Pappers makes its data available to clients through the MCP protocol, which connects a data source to an AI assistant. The reasoning is that many lawyers already use general AIs such as Claude on their cases, which leads to errors on sources: "an AI without sources has no awareness of what it does not know." The opening covers structured data and already vectorized texts that a firm can plug into its own tools to build its workflows. "Our data is worth more when it circulates in the tools our clients already use than when we lock it in an interface."
What's next
Pappers is putting its R&D on agentic work, with the stated aim that "a professional can delegate a complete task, down to the smallest details." Moneyron says three years is short and the team feels it has barely started. It plans about ten hires in 2027.