Private AI and compliance

AI on client data that never leaves your control.

For work where client data can't go to an outside AI provider, we run the models on a server you control and keep a written record of what the AI does and what data it touches.

Enquire now

Who it's for

Firms that hold sensitive client data: legal matters, accounts, insurance submissions and personal details.

Usually one of these has just happened

  • A client contract or your insurer limits where data may go.
  • You had a breach or a near miss and need to tighten up.
  • Your regulator published guidance on AI and you want to follow it.

What directors tell us

What directors tell us

  • “We'd use AI tomorrow if client data stayed in-house.”

  • “I don't know what our staff have already put into ChatGPT.”

  • “Our NDAs say nothing about AI, and that worries me.”

What changes

What changes

  • AI on your sensitive work

    Documents are read and drafted by a model that runs where you choose.

  • A clear answer for clients

    You can tell a client exactly where their data goes and show the record.

  • Fewer surprises

    We find what's already being pasted into outside tools and move that work somewhere safe.

One workflow, for your kind of firm

One workflow, for your kind of firm

  • Law firms

    Search and summarise case files without them leaving your server.

    Measured as hours per file review

  • Insurance underwriting

    Read submissions that hold personal data on your own hardware.

    Measured as submissions handled in-house

  • Accounting and audit

    Read bank statements and payslips locally before they reach your books.

    Measured as documents keyed by hand

We aim to automate the repeatable part of a job, often 60 to 80% of it, and leave the judgement to your people on purpose.

How it runs

How it runs

  1. 01 · Week 1

    Assess

    What data you hold, where it goes now and which rules apply to it.

  2. 02 · Weeks 2 to 4

    Build

    A private model on a server in your office or a cloud account in your name in the EU, connected to your systems.

  3. 03 · At handover

    Document

    A data map and a record of each AI use: what it does, what it touches and who checks it.

How we scope it

How we scope it

The result
The same work, done on hardware you control, with a record showing no client data left it.
What we need from you
Where your data lives today, your IT contact, and a choice between your own server and a private cloud account.
What done looks like
The model runs where you chose, the record is written, and a test shows no client data in outbound traffic.
What a person decides
Who may use it, on which data, and when a case goes to a person.
The price
Fixed and written down before we start. We put a number on what the job costs you now, then price the work so you can see the payback before you say yes.

What it's worth

Put a number on it first

Before we price anything, we work out what the job costs you today: hours a week, times what those hours cost, times working weeks.

Example

Two people spend six hours a week chasing documents. At €45 an hour over 48 weeks, that is about €26,000 a year.

Hours a week12
Cost per hour€45
Working weeks48
A year€25,920

How we do it

How we do it

Models that run where you choose
Open models now handle reading and drafting well. We run them on a machine in your office or in a cloud account you own.
Honest about the trade-off
A private model takes longer to set up and is sometimes less capable than the big cloud ones. We tell you which jobs need it and which don't.
Records written for you
The data map and the AI-use record are the material a data protection impact assessment needs. Legal sign-off stays with your solicitor.

Worth knowing

Worth knowing

  • Ireland's Data Protection Commission says a firm whose staff put personal data into an AI product could be a data controller for it, and should consider a formal risk assessment.

    Data Protection Commission, AI and large language models · Checked 2026-11-12

  • The Law Society of Ireland publishes a template for auditing shadow AI: the tools staff use without the firm knowing.

    Law Society of Ireland · Checked 2026-11-12

Questions

Questions

Enquire now
Is a private model as good as ChatGPT?

For focused jobs like reading documents and filling in forms, often close enough. For open-ended writing the big cloud models are still ahead, and we tell you where that matters.

Do we need our own server?

Not always. A cloud account in your name, in the EU, works for most firms.

Are you lawyers?

No. We build the systems and write the records. Legal advice comes from your solicitor or data protection adviser.

How do you price it?

A fixed price, written down before we start. We first put a number on what the job costs you now, so you can see the payback.