Case Studies

Building private AI for a profession built on confidentiality

Turning sector knowledge, secure infrastructure and human judgement into a daily AI workspace

Sector

Legal and professional services

Client

European legal-tech venture serving civil-law notary offices

Challenge

Bring AI into highly confidential legal workflows without compromising privacy, traceability or professional responsibility

Solution

Private, sector-specific AI platform built around legal knowledge, document workflows, source-based answers and human review

/

/

Building private AI for a profession built on confidentiality

/

/

Building private AI for a profession built on confidentiality

01

The challenge

Generic AI was not good enough for confidential legal work.

The opportunity for AI in legal services is easy to see.

Professionals spend significant time:

  • Reading files

  • Reviewing documents

  • Searching internal knowledge

  • Drafting standard documents

  • Preparing correspondence

  • Checking clauses

  • Summarising cases

Much of this work can be supported by AI.

But the information involved is often highly confidential.

For civil-law notaries, professional secrecy, data protection and accountability are fundamental to the profession.

Sending client files into consumer AI tools was therefore not an acceptable starting point.

The challenge was to create AI that could be useful in daily legal work while keeping professional control, privacy and traceability at the centre.

02

What needed to change

The answer could not be a generic chatbot with legal prompts added on top.

The product needed to understand both sides of the problem.

The technology

  • Private AI infrastructure

  • Secure document access

  • Knowledge retrieval

  • AI agents

  • Workflow automation

  • Permissions

  • Auditability

The profession

  • Legal terminology

  • Professional standards

  • Confidentiality

  • Working practices

  • Document structures

  • Human responsibility

The product needed to bring those two worlds together.

03

Co-created with the profession

Build with users, not for an imagined workflow.

The initial platform was developed together with six civil-law notary offices.

Their teams helped shape the product around work they actually perform every day.

This allowed the product team to move beyond broad ideas about "AI for legal".

Instead, the focus became specific tasks such as:

  • Preparing draft deeds

  • Reviewing files

  • Drafting replies to incoming emails

  • Finding relevant clauses

  • Summarising documents

  • Searching internal knowledge

  • Preparing client communication

Each workflow could then be tested against the realities of professional legal work.

04

What we built

A private AI workspace for legal professionals.

The platform combines a secure AI environment with sector-specific knowledge, assistants and automated workflows.

Employees can use AI to work with documents and knowledge while maintaining visibility over the sources behind the output.

The professional remains responsible for the final result.

1. Private knowledge

Bring company knowledge into a controlled AI environment.

Legal organisations hold large amounts of useful information across:

  • Documents

  • Templates

  • Deeds

  • Internal knowledge

  • Policies

  • Correspondence

  • Case files

  • Clause libraries

The platform creates a secure knowledge environment around this information.

Users can ask questions in natural language and retrieve relevant information without needing to know where every source document is stored.

2. Source-based answers

Legal AI should show its work.

For professional users, an answer alone is not enough.

They need to understand where it came from.

The system therefore provides source references alongside relevant AI output, allowing the professional to check the underlying material before using it.

This creates an important difference between:

AI gives me an answer

and:

AI helps me reach an answer I can verify.

3. Document and file analysis

The platform can support professionals with tasks such as:

  • Summarising files

  • Reviewing complex documents

  • Finding inconsistencies

  • Identifying relevant information

  • Structuring case information

  • Retrieving internal knowledge

The aim is to reduce time spent processing information while keeping the professional in control of the legal interpretation.

4. Document generation

From existing knowledge to a first draft.

Some legal documents follow repeatable structures.

The platform can help prepare first drafts using:

  • Approved templates

  • Known clauses

  • Matter-specific information

  • Internal knowledge

  • User instructions

One example is the preparation of draft deeds.

The product does not remove professional review.

It gives the professional a stronger starting point.

5. Communication workflows

Incoming communication creates substantial administrative work.

The platform can help:

  • Understand an incoming message

  • Retrieve relevant file context

  • Prepare a suggested response

  • Present supporting information

  • Leave the final communication with the professional

Similar workflows can support quotations, document requests and recurring client communication.

6. AI agents and workflows

Move beyond one question at a time.

As the platform developed, the product expanded from a private AI assistant towards specialised agents and workflows.

Current use cases include areas such as:

  • Document generation

  • Communication preparation

  • File summaries

  • Clause assistance

  • Quotation preparation

  • Translation

  • Legal knowledge support

  • Property-transfer workflows

This allows repeatable processes to become structured workflows rather than relying on individual prompting.

05

Human control stays central

AI prepares. Professionals decide.

The legal professional retains responsibility for the work.

That principle shapes the product.

AI can:

  • Search

  • Summarise

  • Compare

  • Draft

  • Prepare

  • Suggest

The professional:

  • Reviews

  • Checks

  • Corrects

  • Approves

  • Advises

  • Decides

This is particularly important in regulated work where efficiency cannot come at the expense of professional responsibility.

06

Privacy by architecture

Confidential work needs a different AI setup.

The platform was designed around private, sovereign infrastructure rather than consumer AI services.

The current deployment uses European infrastructure with isolated customer environments and controls around data access.

The architecture is designed around principles including:

  • Customer data isolation

  • Controlled access

  • Private knowledge environments

  • Audit trails

  • European data residency

  • Role-based permissions

  • Traceability

This makes privacy part of the technical architecture rather than an agreement added afterwards.

07

AI governance is part of the product

Technology alone does not make AI safe to use.

Professional organisations also need clarity around:

  • What information AI can access

  • What employees may use it for

  • What needs human review

  • How activity is recorded

  • How data protection is handled

  • Who owns responsibility

The platform therefore combines technology with support around governance, privacy assessments, adoption and working practices.

08

From tool to daily workspace

The platform started from a simple question:

Can legal professionals use AI without giving up control over confidential information?

It has since grown into a broader workspace covering assistants, knowledge and workflows.

The current product reports:

  • 38+ AI use cases

  • 159 active users

  • 6 offices involved in the original co-creation programme

The platform also reports measurable improvements across standard tasks, review time and employee throughput. These figures should be independently approved before they are carried onto the WeAreBrain website.

09

The result

AI that fits the profession instead of asking the profession to fit AI.

The project created a platform around the requirements of regulated legal work.

Less repetitive work

Routine information processing and drafting can be accelerated.

More usable knowledge

Employees can work with internal documents and professional knowledge through natural language.

Verifiable output

Source references help professionals review AI-generated answers.

Human responsibility

Professional review remains part of the workflow.

Private infrastructure

Sensitive information is handled inside an architecture built for confidential work.

Reusable workflows

Successful use cases can move from individual prompts into shared assistants and automated processes.

10

Why this matters

Regulated AI is a product-design problem as much as a compliance problem.

It is easy to respond to regulation by limiting what technology can do.

A better challenge is:

How do we make AI useful while respecting the rules and responsibilities of the profession?

That requires several things working together:

  • Good infrastructure.

  • Good software.

  • Strong domain knowledge.

  • Clear governance.

  • Human judgement.

This project shows how those pieces can become one product.

11

Relevant WeAreBrain ecosystem

Experts

Domain knowledge is central to the product.

Legal specialists help shape workflows, evaluate output and make sure AI reflects the realities of professional work.

The Forge

The platform is also proof of how deep sector knowledge can become a specialised AI company.

AI services used

Software Engineering

Building the product, AI functionality, workflows and integrations behind the platform.

Data Infrastructure

Creating secure knowledge environments, retrieval and private AI infrastructure.

AI Adoption

Turning privacy, human oversight and responsible use into practical working processes.

Product Design

Designing AI around the daily work of legal professionals rather than around the underlying technology.

Avatar-1
Avatar-2
Avatar-4
Avatar-4
Avatar-3

Building AI for a regulated profession?

Generic AI is rarely enough. We help combine domain knowledge, private infrastructure, product design and human oversight into systems professionals can use with confidence.