AI for law firms and tax advisers: the right file in the right context
Law firms and tax advisers work with documents that must neither be lost nor fall into the wrong hands. Nexoria stores the case files in Nexoria Drive, indexes their contents and makes them findable through semantic document search and the AI assistant Nex, each time with the source passage named. Workspaces keep matters separate from one another, and the same permission check applies to search, Nex and chatbots. Files, index and search run in German data centres, and answer generation runs via Mistral AI. This page describes four tasks from everyday work in a law firm.
Client documents and templates form the starting point. Outlook requires Nexoria Mail; on request, we assess law firm systems based on their interface.
Nexoria Mail is an optional add-on. Email integrations are enabled selectively.
Custom integrations depend on the interface and permissions of your system.
File formats: Doc, Calc and Slides open and save DOCX, XLSX and PPTX. Check rendering, features and the round trip using your own templates.
WHAT THIS LOOKS LIKE DAY TO DAY
Find relevant content in your documents.
Search shared case file contents using specific questions. Use matching passages and the document preview to check the context and read the relevant passages in the original document.
✓Case file contents
✓Search query
✓Source
✓Review
Nex answers a question from your own documents, cites the sources and shows the original document in the preview. Scene from the explainer film, fictitious sample data.Enlarge view ↗
Start written submissions with the context you already have.
Use shared documents for draft wording and revise it in Doc. Legal assessment, completeness and sign-off remain with the responsible professionals.
✓Documents
✓Draft text
✓Professional sign-off
A draft report in Nexoria Doc. Scene from the explainer film with a recreated interface and fictitious sample data.Enlarge view ↗
Separate knowledge collections by area of responsibility.
Organise documents in suitable workspaces and assign roles deliberately. That way, each matter stays in its own knowledge space.
✓Workspace
✓Members
✓Roles
✓Knowledge space
Organise access
AreaReadEdit
Workspace
Members
Roles
Knowledge space
Permissions are preserved
Functional diagram · Sample data
This is Nexoria
Nex: Document search
Document search in Nex: results from documents, emails and meetings. Sample content anonymised for publication.
Product view · Click to enlargeLaw firms: all the detailsPrerequisites, background and frequently asked questions+
The starting point in law firms and tax advisory
A law firm's files are both its working material and its most sensitive point. They grow over the years, consist largely of documents from other senders and contain information that third parties must not see. Taken together, this makes the obvious answer, putting everything into one large searchable repository, the worst solution.
For sectors subject to professional secrecy under Section 203 of the German Criminal Code, such as law firms, tax advisory, healthcare and the public sector, a closed European data flow is often a basic prerequisite. This is a statement about what the sector requires, not about whether Nexoria fulfils the provision. Tax advisory is subject to tax secrecy under Section 102 of the German Fiscal Code (AO), so the requirements for where data is processed are correspondingly high. A statement about what the sector requires, not about whether Nexoria fulfils the provision.
The four tasks below are not categories such as file management, but steps that come up on an ordinary day at a law firm. For each one, you will find the path it takes through the platform and what evidence is available.
Four tasks and their path through Nexoria
Job 1
Find the passage in an extensive expert opinion that mentions a specific deadline, and cite it with its source reference.
Problem
The information sits in one of many client files, often in a PDF several hundred pages long. Full-text search only helps if you know the exact word that appears in the document, and there are several possible wordings for a deadline, a preclusion period or a limitation period.
Document search searches the contents of a workspace semantically and cites the source for every result. Results are shown in a preview with tables, images and formatting, without downloading the document.
Make a scanned old case file searchable so that the person handling it no longer has to leaf through it.
Problem
Old case files exist as scans, in other words as images. As far as full-text search is concerned, they do not exist, and the same goes for AI search. Anyone who needs a detail from 2019 goes to the filing cabinet or clicks through a hundred pages of image files.
Text recognition makes scans and images accessible, and transcription does the same for recordings. The recognised content is indexed and is then available in document search and in Nex.
Store the case files of one matter so that a colleague working on a different matter cannot see them, not even through the AI.
Problem
A knowledge search across the entire collection becomes a problem precisely when it works. If access permissions are only applied after the search, an answer can contain content that is not accessible to the person asking. In a law firm, this is not a matter of convenience.
Strict data isolation per tenant at database and vector store level, a role model with Admin, Member and Viewer, server-side permission checks on every API endpoint.
Draft a written submission from the contents of the case file and be able to trace every statement carried over back to the source file.
Problem
A draft that sounds plausible but whose basis nobody can trace is worthless or dangerous in this profession. That is exactly the well-known weakness of general AI assistants that write without reference to the case file.
Answers from Nex and document search results carry a source reference to the original document. This means every statement carried over into the draft can be traced back to the case file.
Which of these workflows suit your daily work becomes clear with your own content. The components involved are described under Product, and what it costs to use under Pricing. If you would rather see it with your own documents, you can do so via Contact.
Information needed for a professional regulatory review
For each detail, the table gives the source and the date it was checked.
Information on processing, separation and jurisdiction, as requested in a professional regulatory review. As of .
Requirement
Detail
Source
Separation of engagements
Strict data isolation per tenant at database and vector store level, a role model with Admin, Member and Viewer, server-side permission checks on every API endpoint.
Files, the knowledge index and semantic search run on dedicated servers in Germany. Answer generation by Nex uses Mistral AI. Other services and enabled integrations have their own data flows: error monitoring via Sentry, based in the USA, without access to document contents, payment processing via Stripe, and Microsoft Graph and Gmail when the email integration is enabled. For the optional VoiceBot, telephony is provided via Telnyx (headquarters USA, EU branch Ireland). Speech recognition and speech output are handled by Voxtral from Mistral AI. Mistral Large is integrated via Telnyx. The components and providers involved are listed under Security and privacy and in the processing register.
Eight subprocessors, all named in the processing register under Art. 30: Netcup, Hetzner, Stripe Payments Europe, Mistral AI, Telnyx, Functional Software (Sentry), Microsoft Corp. and Google LLC. The figure of eight reflects the status as of 23 August 2026. The current version of the register is authoritative.
In Nexoria's verified Mistral organisation, the use of new API interactions for model training is disabled (account check of 26 September 2026). This setting applies to new interactions processed there. For your own optionally connected AI accounts, the settings of the respective account apply.
For your data protection review, the data processing agreement under Art. 28, the processing register under Art. 30 and the data protection impact assessment under Art. 35 GDPR are publicly available. The details on this page are Nexoria's own statements; the data flows for each processing step are described in the privacy policy and the processing register.
Together, we review your data sources, workflows and requirements. Get to know Nexoria through your specific use case and plan the right way to get started with our team.
Frequently asked questions
Does Nexoria comply with Section 203 of the German Criminal Code?
Compliance with professional secrecy is the responsibility of the professional bound by it, not of the software provider. Section 203(3) of the German Criminal Code also requires a contractual obligation for the persons involved. Whether and in what form Nexoria offers this obligation has not been published. What Nexoria describes is what the sector requires: for sectors subject to professional secrecy under Section 203 of the German Criminal Code, such as law firms, tax advisory, healthcare and the public sector, a closed European data flow is often a basic prerequisite. This is a statement about what the sector requires, not about whether Nexoria fulfils the provision.
Are the files of different matters kept separate from one another?
Yes, through workspaces. Each workspace has its own knowledge base, and the same permissions apply to document search, Nex and chatbots. Strict data isolation per tenant at database and vector store level, a role model with Admin, Member and Viewer, server-side permission checks on every API endpoint.
Where are the case files stored and where does the AI run?
Files, the knowledge index and semantic search run on dedicated servers in Germany. Answer generation by Nex uses Mistral AI. Document embeddings and semantic search run on servers in Germany, using models that Nexoria operates there itself. Files, the knowledge index and semantic search run on dedicated servers in Germany. Answer generation by Nex uses Mistral AI. Other services and enabled integrations have their own data flows: error monitoring via Sentry, based in the USA, without access to document contents, payment processing via Stripe, and Microsoft Graph and Gmail when the email integration is enabled. For the optional VoiceBot, telephony is provided via Telnyx (headquarters USA, EU branch Ireland). Speech recognition and speech output are handled by Voxtral from Mistral AI. Mistral Large is integrated via Telnyx. The components and providers involved are listed under Security and privacy and in the processing register.
Are case file contents used to train AI models?
In Nexoria's verified Mistral organisation, the use of new API interactions for model training is disabled (account check of 26 September 2026). This setting applies to new interactions processed there. For your own optionally connected AI accounts, the settings of the respective account apply.
How do I review a draft written submission created by the AI?
Nexoria generates drafts from the content shared in the workspace and cites the source document for every statement carried over. This lets you check each statement directly against the case file before you finalise the draft in Doc. As usual, professional responsibility lies with the qualified professional.
Which documents support our security review?
The security overview documents technical and organisational measures, data processing and service providers. It supports the review of your requirements before rollout.
Sources
As of
German Criminal Code (Strafgesetzbuch), Section 203: Violation of private secrets. Federal Ministry of Justice, Gesetze im Internet, 1998. https://www.gesetze-im-internet.de/stgb/__203.htmlRetrieved on .Source for professional secrecy and for the obligation of persons involved required under subsection 3.
German Fiscal Code (Abgabenordnung), Section 102: Refusal to provide information to protect certain professional secrets. Federal Ministry of Justice, Gesetze im Internet, 2002. https://www.gesetze-im-internet.de/ao_1977/__102.htmlRetrieved on .Source for the tax secrecy mentioned in the text.
Mistral AI, Legal notice. Mistral AI SAS, 2026. https://legal.mistral.ai/legal-noticeRetrieved on .Source for the legal form and registered office of the model provider that handles answer generation.
As of
Next step
The most reliable test is your own case file. In a conversation, we clarify how matters are divided across workspaces, how old case files are made accessible and what indexing your existing files costs.
