The challenge
Daily work was distributed across documents, email, calendars, handwritten notes and external systems. Information about one matter lived in several places, and keeping it complete required time and detailed knowledge of the team’s working methods.
Large proceedings created an additional challenge. Evidence could span many files and hundreds of case-file volumes in different formats and versions. Preparing that material for reliable search and analysis was a substantial process in itself.
- organise cases, documents, tasks and billing
- reduce manual retrieval and re-entry of information
- work efficiently with very large document collections
- use language models without losing control of the sources
- preserve the exact document set and version used for an analysis
The solution
LexCoreAI was designed and delivered as the central workspace for the legal practice. It is not a separate chatbot attached to a file repository: matter management, documents and AI analysis form one controlled workflow.
The platform synchronises information with the Polish Courts Information Portal. Retrieved documents are registered, processed and assigned to the correct matter, reducing manual checks and data transfer.
- cases, clients, contacts and proceeding participants
- document repository with folders, versions and access control
- tasks, activities, time recording and team calendars
- costs, rates and multiple billing models
- document generation and reusable templates
- live progress and audit information for long-running operations
Preparing case files for analysis
Documents are converted, recognised and organised in stages before a language model can use them. The system keeps each text fragment connected to its source file and page, and creates a defined version of the material used for analysis.
Adding documents later does not silently change an earlier result. Every answer remains linked to the exact version of the case material on which it was based.
- receive and classify case documents
- convert files and perform OCR
- retain document and page provenance
- build a versioned analysis set
- run and record the analysis
- return an answer with source references
A source-grounded AI assistant
The assistant answers questions about case documents, helps structure the material and prepares working drafts. It can identify the documents and pages used to reach a conclusion and flag content that requires additional verification.
Two models may cooperate: one prepares an answer and another reviews it against defined criteria. The final legal assessment always remains with the lawyer. AI reduces the effort of working with the material without replacing professional responsibility.
Key capabilities
- one view of cases, participants, documents, tasks and deadlines
- versioned case files and role-based access control
- court portal synchronisation, document conversion and OCR
- analysis of large, controlled document collections
- answers linked to documents and page numbers
- controlled cooperation between multiple AI models
- Word documents, templates, costs and time records
- live operation progress, automated tests and separate environments
Business outcome
LexCoreAI brought the legal practice’s key processes into one application. Users can find information faster, monitor deadlines, organise documents and prepare material for analysis.
Automation reduces work related to data retrieval, document conversion, source preparation and search. At the same time, AI results remain verifiable against specific documents, pages and case-file versions.
Technologies
- C# and ASP.NET Core MVC
- Microsoft SQL Server
- Python and FastAPI
- Docker and Portainer
- SignalR and Server-Sent Events
- Google Document AI OCR
- Polish Courts Information Portal API
- Claude and OpenAI
- OpenAI Codex and internal MCP tools
- Automated .NET and Python tests
- Separate DEV and PROD environments