An assistant that asks back — and a second mind.
The AI assistant works with your case relationship graph — switchable via „Include relationship graph of the case“ — as well as model tiers and attachments, and asks back when something is unclear instead of guessing. In the Lex & Lexi debate, Lex and Lexi provide two voices: argument and counter-argument.
Ask back instead of hallucinating
When a request is ambiguous, the assistant asks targeted questions before drafting — so no false assumptions arise. With case context, model tiers (Auto/Quick/Standard/Premium) and attachments.
- Targeted follow-up questions when unclear.
- Case context and attachments per thread — with an outline, a status recap for the case memo, branches for variants and projects for several cases.
- Edit Word files by instruction — optionally as suggested changes, with redline and clean versions; match exhibit references against the directory. You confirm every assignment. For fine work by hand, you open the same file in Canvas.
- The assistant can search facts already extracted from the case file at no AI-quota cost — only when that isn't enough does it read the original document again, which does use quota.
Before I draft: is receipt of the termination on 14 Mar undisputed, or are we contesting the date? That changes the line of defense.
Argument and counter-argument
Two AI voices debate your case: Lex formal and procedural, Lexi practical and unconventional. So you see the main line and the risks before the court does.
- Hybrid debate in several modes (analysis, strategy …).
- From the debate straight into a pleading/contract — with provenance.
Formally: the claim should be dismissed for lack of conclusive pleading on receipt.
Risky — the court could assume receipt via prima facie evidence. Second line: prepare a severance settlement in the alternative.
Agreed: primary motion to dismiss, signal willingness to settle in the alternative.
Sourced, not invented
On request, the assistant backs its answer with sources and lists references. Whatever is not verifiable is marked as such.
- References with a source instead of free-floating claims.
- Unverifiable items are flagged, not asserted.
- § 4 KSchG — statute confirmed
- § 130 BGB — statute confirmed
- BAG on deemed receipt — check the reference before use
Agents in this workflow
AI assistant with persistent threads, case context, model tiers, attachments, branches and projects, editing of Word files by instruction (every change a new version) and export as Markdown, PDF or Word.
Lex & Lexi debate: two agents in several modes — transferable straight into a pleading/contract.
A second opinion like an experienced colleague — a structured assessment of strategy and risks.
Control & safety
- The assistant asks back instead of inventing assumptions.
- Answers are sourceable; unverifiable items are marked.
- Case context stays client-separated (RLS).
- EU hosting, multilingual. Depending on your firm's setting, the interactive assistant masks personal data before sending it to the model; the preview shows what is masked, and you choose clear text deliberately per request, chat or case – unless your firm blocks it.

Let your case files think along.
The AI lawyer portal — EU-hosted, profession-compliant, client-separated. You keep control, the AI does the groundwork.
Video and images
1:06For law firmsThe AI works your way: per case file, per firm
1:09For law firmsAI Assistant: A Legal Question on the Case File, Answered Live
1:10For law firmsMasking made visible: what the AI receives of your client data
More images: Firm & DMS →More images: AI assistant →Videos →Screenshots →

