An Australian economics consultancy wanted the productivity of tools like ChatGPT, Claude and Copilot, without losing control of its data or how AI was used. We built a custom LLM harness: its own AI assistant, with governance and the firm's workflows built in.
About Our Client
Overview
An Australian economics consultancy working with sensitive client data and complex economic analysis.
The Challenge
The firm's consultants work with sensitive client data and complex economic analysis. Off-the-shelf AI tools promised big productivity gains, but raised concerns: where client data would go, how outputs would be checked, and how to keep AI use consistent across the team. The firm wanted AI tailored to how it actually works, not generic prompts.
The Approach
-
Built a custom LLM harness. A private AI assistant giving consultants the capability of leading models, under the firm's own control.
-
Built governance in. Access controls, usage rules and oversight are part of the platform, not an afterthought. Data staying within their environment was critical.
-
Embedded the firm's workflows. Common tasks such as research summaries, report drafting, data analysis run as built-in workflows, so every consultant gets consistent, high-quality results.
The Result
- AI productivity without handing client data to a public tool
- Consistent, governed AI use across the firm
- A platform the firm owns and can extend as its needs grow
