AI case intelligence for UK law firms

Ask your case files anything.

Bibliolegis reads every document in every matter, answers your questions in plain English and shows you the exact passage each answer came from. Securely, and only for the people allowed to see it.

  • Every answer cited
  • Access set per matter
  • Every question audited
Mackenzie v NewmanExample, names invented
When was Newman first interviewed and what did he say about the van?

He was interviewed under caution on 14 March 2025 at Brighton custody centre1. He said the van was parked outside his brother's flat and he hadn't driven it that week2.

Record of interview, page 7
1Custody record, page 22Record of interview, page 7

The hidden cost

Sarah is chasing a witness statement that should be in the system. Mark is rebuilding a report from three different places. Somebody is reading a 200 page bundle for the one line that matters.

The real cost of scattered case files isn't the licence fee. It's the hours spent looking instead of billing.

You need to know how many burglary matters came in this quarter. Instead of three systems, two spreadsheets and a prayer, you ask.

53 open matters, all shouting for attention. Ask which ones have a hearing this month and get the answer with the documents behind it.

Summarised in seconds, with every point linked back to the page it came from so you can check it before you rely on it.

What it does

Chat with your matters

Ask about one matter or everything the firm holds. Follow up the way you would with a colleague, it remembers what you were talking about.

Answers you can check

Every claim carries a numbered citation and the exact passage it rests on. If nothing in your files answers the question, it says so rather than guessing.

Drop in files, get matters back

Drag in a pile of case files. Bibliolegis works out which case each belongs to, names the matter and pulls out the parties, court, reference and key dates.

Reads scanned documents too

PDFs, Word files and old scanned bundles. Pages without text are read by OCR, so the exhibit someone photocopied in 2019 is searchable too.

Your choice of AI model

Pick the model that answers you, from OpenAI or Anthropic's Claude. The firm decides which ones are on offer.

A workspace for every matter

Files, chronology, hearings, parties, notes and research in one place, with pinned matters a click away.

Questions that need counting, answered

“What's the typical sentence for shoplifting in Brighton?” isn't something a search box can answer. Bibliolegis reads the offence, court, date and sentence out of every case as it arrives, then counts across them. You get a breakdown by outcome and the number of cases behind it, never a bare average. When the sample is too small to mean much, it tells you.

Shoplifting, Brighton Magistrates', last 3 years

42 cases · example data

Community order19
Fine12
Suspended sentence7
Immediate custody4

Immediate custody: median 4 months, from 4 cases. Too few to call typical.

What an hour is worth

Work it out with your own figures. The hour spent hunting through bundles is an hour nobody can bill.

10
£200
1 hour

Billable time back each year

£44,000 per fee earner

An illustration from the numbers you enter, based on 220 working days a year. Not a promise of what any firm will save.

Stone columns of a classical court building

Security and confidentiality

Client files carry your duty of confidentiality with them. Bibliolegis was designed around that from the first line of code, with UK GDPR in mind.

Access set per matter

Colleagues only see the matters they're on. A file someone can't open never appears in their search results or shapes their answers.

Every question on the record

Who asked what, when, which model answered and which documents it read. An audit trail your COLP can actually use.

Encrypted throughout

Files and data encrypted in transit and at rest, in a workspace that belongs to your firm alone.

Never used to train AI

Your case files are sent to the model only to answer your question. They aren't used to train it.

How it works

  1. Drop in your files

    PDF, Word or scanned. As many as you like, a matter at a time or the whole archive.

  2. Bibliolegis reads them

    Text is pulled out, scans are read, and the key facts of each case are filed into matters you can open.

  3. Ask, and check the source

    Type a question and get an answer with the passages behind it, one click from the original page.

On the way

Coming soon

Ask out loud

Hold a button between court and the office, ask about a matter and hear the answer, with the sources left on screen.

Coming soon

Your mailbox and diary

Connect Outlook or Gmail and your questions reach the correspondence too. Private to you, never shared with colleagues.

Coming soon

A morning brief

A few lines before the day starts: who you're meeting, what's in court and what's due, each linked to where it came from.

Coming soon

Press watch

Coverage of your matters' parties and cases from national, legal and local titles, gathered for you.

Questions

Is this a case management system?+

No. Bibliolegis doesn't do time recording, billing or client accounts. It sits alongside whatever you use for those and makes the documents themselves searchable and answerable.

What if the AI gets something wrong?+

Every answer shows the passages it relied on, so you can check it in seconds. When nothing in your files supports an answer, Bibliolegis says it couldn't find one rather than inventing it.

Which files can I upload?+

PDF and Word documents, including scanned PDFs, which are read page by page.

Can everyone in the firm see every file?+

Only if you want them to. Each matter has its own members, and a document in a matter someone isn't on stays invisible to them, in the file list and in every answer.

Do the AI providers keep or train on our files?+

Your files are sent to the model you choose only to answer the question you asked, and aren't used for training.

Upload a matter and ask it a question. It takes a few minutes.