AI for law firms — with professional secrecy

Legal answers based solely on sources

Knowledge and documents

Problem: Invented case references and non-existent judgments in material prepared by AI

How it works: Answers only from the central RAG knowledge base, from a selected set of case files and case law; source preview; the Knowledge Gap Register flags a missing basis

What you gain: Every claim has a source you can click, and a missing judgment in the base is a message, not an invented case reference

Assistant for drafting pleadings

Process automation

Problem: Every pleading is written from scratch

How it works: A case workspace with the files, timeline and correspondence; a local agent reads Outlook mail without copying it to the central base and prepares a draft pleading for review

What you gain: The draft is built from the case files and correspondence, so the lawyer starts with a substantive review, not a blank page

Automatic tracking of procedural deadlines

Process automation

Problem: Procedural deadlines are tracked by hand in several places at once

How it works: Schedules, a “Morning Triage” that sets the day’s priorities, and a Kanban task manager with dates

What you gain: One view of deadlines for all cases, prioritised for today — the risk of missing a deadline no longer depends on a single calendar

Automatic time reporting

Analytics and reporting

Problem: Time recording for cases is late and inaccurate

How it works: Automatic timesheets based on system activity and the calendar

What you gain: Time spent on a case is recorded as you go — the client invoice is complete, not reconstructed from memory

A confidential AI environment for case files

Security and sovereignty

Problem: Professional secrecy rules out processing case files in commercial chatbots

How it works: Confidentiality mode and permanent marking of case files as confidential; confidential data blocked from routing to cloud models; deployment on-premises in the firm’s infrastructure

What you gain: The line between a general question and case data is guarded by the architecture, not by the lawyer’s attention “mid-sentence”

Documented use of AI in cases

Compliance and oversight

Problem: The code of ethics requires documenting when and how AI tools were used in a case

How it works: An immutable log of model use; AI Disclosure Injection that marks AI-generated content; an AI use policy and an agent register (CDF-F2)

What you gain: The firm can show the client and the bar council which tool processed the case data, when and why

Does data leave the EEA?

Not unless the firm decides so. TWIN:DESK runs on-premises or in a cloud with data located in Poland or the EU, and PROXY:AI lets you allow only models that run locally or in the EU

Will our data be used to train models?

No, if the firm uses local models or providers whose contracts exclude training on client data. PROXY:AI lets you allow only such models and masks personal data before a query is sent

Does the tool provide adequate protection within the meaning of § 23e?

The assessment is up to the firm. SAIE gives it the basis for one: control over data location, case-level permissions, a log of who processed documents and when, and documentation to demonstrate due diligence