Predictice: our review

Predictice starts from a different reflex than classic legal research. Instead of returning a list of decisions to read one by one, the tool turns them into a figure. For a given head of claim — dismissal without real and serious cause, abrupt termination of a business relationship, personal injury — it calculates an acceptance rate, a rejection rate and an average compensation amount, drawn from tens of thousands of comparable decisions. "Predictive justice" gave the company its name in 2016 and still structures the entire product.

Putting a number on it before reading

In practice, a lawyer enters the head of claim, filters by jurisdiction, date and the compensation amount sought, and gets a summary: how many comparable decisions, what proportion upheld the claim, what average amount was awarded. Each decision remains available in an enriched reading view, with a case timeline and an interactive outline. Around this litigation core, Predictice has added annotated texts (codes, collective agreements, international tax and social-security treaties) and desk tools: document audit, version comparison, translation via DeepL Pro.

An entry price few others publish

In a sector where Doctrine and Case Law Analytics systematically point back to a quote, Predictice publishes a price: €250 excl. VAT per month for the Essentiel plan, on an annual commitment. A sole practitioner can therefore compare this figure against their budget before even filling in a contact form. Hosting follows the same logic of transparency: OVH SAS, in Roubaix, with data stored within the European Union according to the legal notice and privacy policy — a point one can check rather than take at face value. It's the argument that matters for a firm bound by professional secrecy; see our page on AI for the legal sector.

What transparency doesn't cover

The published price stops at individual use: as soon as a firm wants to equip several staff, the "Team" plan goes back to a quote, as everywhere else. More troubling from a confidentiality angle, none of the pages reviewed — privacy policy, terms and conditions, legal notice — say whether documents uploaded to the Assistant or the Library are used to train a model. Doctrine and Ordalie publish an explicit commitment on this point; Predictice publishes none, which leaves the question open. The privacy policy also acknowledges a transfer of certain data to subcontractors outside the European Union, in the United States, relying on the Privacy Shield — a legally fragile mechanism since 2020, even though the clause concerns cookies rather than case files.

Our recommendation

Predictice suits a lawyer or in-house counsel who litigates often and wants a risk figure before starting proceedings, with a clear budget from the outset. It suits less a firm that wants, on top of the price, a written commitment on the fate of its documents in relation to AI: ask for it in writing before signing. A prior audit and advisory engagement helps frame this question before the seven-day free trial.