Your team of 7 AI agents, each with one job
Each is a separate Codex agent. None of them can approve.
Manager
Plans and assigns the work
Knowledge custodian
Keeps the sources and page references
Document architect
Plans the smallest set of documents
Writers
Draft each document
Assessment reviewer
Checks against the rating
Evidence verifier
Checks every claim has a source
Editor & publisher
Finalises and versions
Gather your rulebook
Collect the government handbooks and your plans in one place.
Map what’s needed
List every requirement for each of the 81 capabilities.
AI writes the documents
AI drafts each policy, procedure and template to match the handbooks.
AI double-checks
Two separate AI reviewers check every draft before you see it.
You approve
Nothing is final until your team signs off.
Track progress
A simple dashboard shows what’s done, what’s waiting, and what needs your team.
We start small: AI Governance first, proven on your own tests
The pilot is proven when each of these deliberate test cases stops a document from being released:
We also show that work can pick up after a stop, will not run twice by mistake, and is checked again if a source changes.
To get started, I’ll just need
- DGE Handbooks v1.2
- Reworked uplift plan
- Evidence taxonomy
- Oct–Nov evidence schedule
- In-scope capabilities and target ratings
- Access to approved hosting
- Existing ADIO policies and approved records
Whatever scope you need, from one pillar to all seven, I can deliver it.
How it runs on Codex
Codex is the AI tool that does the work. Tap any step to see it in action. Your team stays in control.