Something has changed in legal research. Firms that renewed the same database contract for a decade are running evaluations. Managing partners are sitting in demos. And the money has followed: legal AI companies are raising nine-figure rounds at multi-billion valuations. This is good news. It proves lawyers want something new.

If you're reading this and not in the legal world, you probably picture legal research the way TV shows it. It's 2 a.m. A junior associate flips through page 4,000 of something. Then, with the dramatic music swelling, they shout "I found it!" One tiny detail in a footnote, and the whole case turns.

The TV version isn't that far off. Lawyers and all legal professionals really have spent whole nights hunting for the one sentence that matters. The difference now is that it can take seconds, and there’s a real demand for that.

All that demand means the market is about to get louder. More vendors, more stands, more demos. You wouldn't believe how many partners tell me they've sat through several legal AI demos and couldn't say how they differed. From the outside, it all sounds the same.

And I have a strong feeling about why.

The model is a rental

Almost every legal AI company licenses its intelligence from the same handful of labs. The lead of the best model lasts months, then everyone catches up. That's wonderful for lawyers, because good reasoning is becoming cheap and everywhere. But it means the model can't be what sets a product apart for long.

What doesn't become a commodity is the data underneath. A model you license is a rental. A corpus you build is an asset. That's where I believe this space will be won, and where it can be at its best.

Becoming data first

Only around 3.5% of German court decisions are ever published. The other 96.5% still decided real cases, real jobs, real businesses. Most legal professionals will never access them.

Across Europe, most case law sits in scattered databases or isn't public at all. The next big leap in legal research won't come from a smarter model. It will come from bringing that hidden law into the light. We've already had customers ask whether a verdict they found with us was real, because it existed nowhere else. That's the "I found it!" moment (minus the all-nighter).

A judgment means more when you can see the appeal that reversed it. A statute means more with its preparatory works attached to the exact section, and the EU directive it implements one click away. This patient, unglamorous work is what lets AI reason like a lawyer does, following the hierarchy of sources instead of guessing.

Let’s also not forget that many European lawyers don't work in one jurisdiction. A German contract, a Swiss parent and a Danish customer is an ordinary Tuesday. Tools that do one country well and the next one thinly will feel as outdated in five years as boolean search does now. These lawyers shouldn't have to log into one more app to do their work. The most valuable legal data will be the data that reaches lawyers wherever they already work: in Word, in their own workspace, in the tools their firm builds.

What this means if you're choosing now

Ignore the funding headlines, ours included. Ask every vendor three questions: what is your data, and where does it come from? Can I check every answer in a few clicks? And will the data still be current in five years, when today's model has changed?

That's where we've spent most of our years at Pandektes: more than two million decisions, legislation and preparatory works across Europe’s key jurisdictions. We're spending our energy on doing the absolute best job on the data. So now, lawyers can spend their nights on judgement, not flipping through 4,000 pages.