Practice areas

Criminal law

Voluminous files and very short deadlines

A criminal file can run to thousands of pages, with deadlines of a few days. Ulpian classifies the documents, builds the chronology and prepares the first read, so the lawyer can spend their time on the defence.

What you will see in Ulpian

Fictitious matter, sample data.

Facts with their source

Each fact is attached to the document it comes from. Click a fact to read the source passage; contradictions are flagged.

Contract signed12 JanFirst delivery3 FebBuyer's complaint28 FebExpert report14 MarFormal notice2 AprTermination20 Apr

Choose a fact on the timeline.

What weighs on the work today

  • 01

    Thousands of pages

    Interview records, reports and seized documents arrive in volumes nobody can read in full before the deadline.

  • 02

    Deadlines of a few days

    Appeals and observations must go out quickly, often while other hearings are being prepared.

  • 03

    Absolute confidentiality

    The file must be seen only by the defence team.

A matter, from first contact to invoice

  1. Step 0101

    The matter opens, walled off for the defence team.

  2. Step 0202

    The file's documents are added, classified and dated; facts join the chronology.

  3. Step 0303

    The first read flags useful passages and contradictions between statements.

  4. Step 0404

    Deadlines are confirmed, submissions prepared, and the hearing runs from the preparation sheet.

What the AI prepares

  • The first read of documents, with the source passage of each element flagged.
  • Contradictions between statements, attached to their documents.
  • Research in public case law, with verified citations.

What stays with you

  • The defence strategy.
  • Every deadline and every filing.
  • Every exchange with the client and the authorities.

Frequently asked questions

Are the documents used to train the AI?

No. Your documents and queries are not used to train any model. Before a call to an external model, identifiers can be replaced by placeholder tokens.

Can we use a local AI model?

Yes. You can use your own keys or a local model, and impose a processing region. Every call is logged.

See it on a matter that looks like yours

The diagnostic shows you in three minutes where your firm loses time. The call shows you Ulpian on a typical matter, in twenty-five minutes.