AI for law firms: client files never leave the building.

Your lawyers already know what AI can do; many of them are quietly using it. The problem was never capability. It is that a firm's stock in trade is privileged, confidential material, and the standard AI tools all work by transmitting that material to someone else's infrastructure. An Ermos AI unit lives in your office, so the capability arrives and the client files never leave. Our current unit is Dominion.

What your team does with it

Contract and lease review

Upload an agreement and ask what is unusual against your own precedents: notice periods, indemnities, missing clauses, defined-term inconsistencies.

Discovery and brief summarisation

Condense large document sets into chronologies and issue summaries your team verifies, instead of reading cold.

Precedent-grounded drafting

First drafts of advices, letters and standard documents in your house style, drawing on your own precedent bank rather than the open internet.

File chronology

Reconstruct the timeline across correspondence, file notes and drafts on a matter, with each event traceable to its source document.

Letter and correspondence drafting

Client updates and routine correspondence drafted from the file, ready for review rather than from scratch.

Institutional memory

Ask what the firm has previously advised on a point and get answers grounded in your own archive, with citations to the documents.

The compliance frame, plainly

Two obligations do the work here. Legal professional privilege protects confidential communications made for the dominant purpose of legal advice or litigation; the live question modern AI raises is whether routing client material through a third-party service sits comfortably with the confidentiality that privilege depends on. And rule 9 of the Australian Solicitors' Conduct Rules binds practitioners to client confidentiality subject to narrow exceptions. Neither rule says "no AI". Both make the deployment posture the question that matters: with a no-egress unit, there is no transmission to a third party to analyse, because there is no transmission at all. For firms that run entirely in the cloud with no file server, Ermos Edge is hosted for you in Australia.

Cloud AI vendors answer with training policies and certifications, and the good ones mean it. Our Claude and ChatGPT Enterprise comparisons set out honestly what those products get right, and why "not trained on" is a different promise from "never left the building".

Built for how a firm actually runs

Your Ermos unit is installed on your network and fully managed: monitoring, model updates and support are included, and ErmosIQ tunes the models to your precedents, your matter types and your house style over time. Partners and staff use it like any modern AI assistant. Nothing requires your IT support beyond a power point and a network port.

The use case library

Process-level guides to running specific workflows on an Ermos unit, each on its own page. We add to this library as clients show us new ways they work.

Dominion is priced per unit, per month on a 12-month term, sized to your practice: unlimited users, no per-user licences, no usage charges. Exact pricing is confirmed in the scoping consultation.

Common questions

Can law firms use ChatGPT for client work?
Many firms permit general research and drafting with no client information, and prohibit client material in cloud AI tools. The difficulty is enforcement: staff paste from files because it is useful. An on-premise unit resolves the tension by making the useful path and the compliant path the same path.
Does using AI risk waiving privilege?
Privilege turns on confidentiality, so the careful question is whether disclosure to a third-party service is consistent with maintaining it. That analysis is contested for cloud tools and does not arise for a no-egress unit, where no third party receives anything.
Will it work with our document management system?
The unit indexes the documents you give it access to on your network, including Microsoft 365 stores. Scoping covers your DMS during the consultation.
Is a local model good enough for legal work?
For open-ended general reasoning, frontier cloud models lead. For work grounded in your own matters, precedents and correspondence, a tuned local model answers from material no cloud model has seen. Most fee-earning work is the second kind.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo
Keep reading
On-premise vs cloud AI What on-premise AI costs The Dominion AI unit How ErmosIQ learns your practice