Private AI Deployment for Public Infrastructure
Designed and deployed a private AI workspace with full decision logging and custody chains for a civic institution managing $2B+ in assets.
National agency coordinating multi-region infrastructure programs
The Challenge
Critical infrastructure decisions were fragmented across hundreds of email threads, PowerPoint decks, and informal chats. When audit time came, teams spent weeks reconstructing decision rationale.
There was no consistent language for talking about approvals. One region's 'sign-off' was another's 'endorsement.' Funding changes happened without documented authority.
Staff had started using public AI tools to draft reports, creating data-leak risks. The organisation had no visibility into what documents were leaving the building.
Our Approach
We started with a three-week discovery, mapping the actual decision flows across regions. Not the org chart—the real flows. Who actually decides what, and where does that information go?
From this, we designed a canonical decision model: every infrastructure decision has an authority, a scope, an evidence base, and a lifespan. This became the foundation of the ledger.
We deployed a private AI workspace using local RAG. Documents never left the organisation's servers. Every AI response included inline citations back to source documents.
We implemented PATTERNS modules for common decision types: approvals, risk escalations, funding amendments. Each module contained the specific fields and workflows needed for that decision type.
“For the first time, we can see the full lineage of any decision. When an auditor asks 'why was this approved?' the answer is two clicks away, not two weeks of searching.
Deliverables & Outcomes
What we delivered
- Private AI workspace with local RAG and citation tracking
- Decision record (ledger) with lineage
- Role and ownership map for approvals
- Interface standards and PATTERNS modules
- Audit and handover documentation
Measured results
100%
Decision Coverage
of approvals captured in ledger within 90 days
2 hrs
Audit Response
average time to answer audit queries (was weeks)
Zero
Data Leak Risk
all AI queries remained on-premises
Project Gallery
The ledger is now being extended to procurement decisions and the private AI workspace is being rolled out to three additional departments.
Related Services
See how the protocol works
Each case study demonstrates how Respectful Mediation, Structured Cultivation, and Natural Harmony operate at scale.