PL EN
// Case study · law firm system · AI

A litigation document system for a law firm

We designed an on-premise system for a law firm handling construction and infrastructure disputes, public procurement appeals and commercial litigation. An AI model reads the case files, extracts findings with the document and page they come from, spots contradictions and drafts pleadings — while every decision stays with the lawyer.

See the prototype ↗
161documents in the prototype case files
5,043articles in the law base from the Sejm ELI API
5 stepsfrom pleading type to a finished .docx
2interfaces: lawyer and administrator
// What we did

Case files, findings and pleadings in one system

Lawyers no longer retype dates and amounts from the files — they get them with a link to the scan and decide what to do with them.

// 01

Case files and scan recognition

Scans of pleadings, contracts and protocols are recognised and each page is scored for quality. Pages below the threshold — a stamp over text, a copy of a copy, a handwritten note — go to a review queue and never feed into pleadings.

// 02

Findings with their source

Dates, amounts, claims, deadlines and contract parties are extracted from documents. Every field has a document, a page and a confidence score, and one click highlights the spot on the scan.

// 03

Discrepancies and claims schedule

The system compares findings across documents and shows contradictions side by side — but it does not decide; the lawyer does. The claims schedule puts the firm’s position next to the other side’s.

// 04

AI pleading builder

Five steps: pleading type (only those that fit the case stage), findings, sources, argument blocks and file assembly. The firm’s template fills in the parties and relief sought, the model writes the argument, and every number links to the scan.

// 05

Deadlines and law base

Procedural deadlines — appeal, cassation, KIO, public procurement court — with the legal basis and the date they run from. Plus 5 statutes in consolidated text pulled from the Sejm ELI API.

// 06

Search across the firm’s own pleadings

A substantive question, e.g. “how did we argue that works fell outside the lump sum”, searches the firm’s pleadings, cases, documents, statutes and case law — only within what the user is allowed to see.

// Screens

What the system looks like

Screens from the interface prototype. Cases, companies, amounts and case numbers are fictional.

// Security

The firm’s data stays with the firm

Court files are sensitive data, so the system is designed to run on-premise with full access control.

  1. 01On-premise, in the firm’s own infrastructure.
  2. 02Accounts, roles and 2FA; case access by team, with a “see as” preview.
  3. 03Audit log: who opened and changed what.
  4. 04Versioned pleading templates and model instructions with change history.
  5. 05Backups with restore testing and a data handover package.

Want a similar system in your firm? AI for law firms →

// Let’s talk

Want AI to read the case files for you?

Free online consultation. We will show you which stages of drafting can be sped up in your firm.

info@axisway.com+48 791 793 584ul. Szewska 8, 50-122 Wrocław