ned benjamin

Australian law firm · AI consulting

Law firm AI setup

Two pieces of work for one firm. I set up its AI system, which is now used firm-wide. Separately, I wrote a proposal for keeping client data in-house when staff use cloud AI models.

Outcome
An AI system integrated with the firm’s legal software, now used firm-wide.
Role
AI consultant: set-up and privacy advisory
Team
The firm’s principal
Timeline
2026 – present
Methods
ChatGPT projects, skills and connectors · usage rules · legal software integration · privacy proposal
Deliverables
AI system set-up (projects, skills, connectors, usage rules) · anonymisation layer proposal
Skills
AI systemsPrivacyLegal tech

Delivered: the firm’s AI system

I set up the firm’s AI system on ChatGPT and integrated it with the firm’s legal software. It is now used firm-wide.

  • Projects

    Shared context for the model: the firm’s writing style, document formats and more.

  • Skills

    Built from that same context, plus skills designed specially for the firm’s work.

  • Connectors

    Links to the firm’s legal software.

  • Usage rules

    How staff use AI with client work.

Advised: a local anonymisation layer Proposed, not built

The firm’s principal was uncomfortable sending client data to cloud AI models. I wrote an advisory proposal for a local anonymisation layer: software on the firm’s own hardware that would strip client details before any data reached a cloud model. This was a proposal only; it has not been built.

Proposed data flow Proposed

  1. Inside the firm, on in-house hardware

    1. Staff requestMay include client details
    2. Local anonymisation layerStrips client details
  2. Cloud AI modelReceives anonymised data only
In the proposal, client details never leave the firm’s hardware.

Working with engineering

The system connects to the firm’s legal software through connectors. The anonymisation proposal placed a local layer on the firm’s own hardware, in front of the cloud models (see the proposed data flow).

Reflection

The set-up covered how staff use AI, not just the tools themselves. The privacy proposal took the principal’s concern about client data seriously and designed around it, rather than asking the firm to simply trust the cloud.