E-book
Governing AI Agents in the Public Sector
A working guide for the people who sign off on an AI agent in a Canadian department or agency, from the programme owner to the CIO. About ten pages, in English and French.
What is inside
Eight chapters, each ending in a checklist
- How to sort every agent task into drafting, deciding or acting
- What the Treasury Board Directive on Automated Decision-Making asks of you, and when to start the Algorithmic Impact Assessment
- Data, privacy and hosting questions to settle before the agent reads a single record
- Human review that catches errors, and how to measure it
- Testing before launch and watching after, in both official languages
- The logs and records an auditor or an access to information request will ask for
- Supplier and contract questions for an agent built on someone else’s model
- A first ninety days for a pilot, with a stop rule
Who it is for
Written for the people who sign off
Programme owners who want an agent and need to explain it. Privacy and security officers who must be comfortable with it. The CIO who answers for it. It is a practical guide, not legal advice, and it tells you to check the current policy text with your own advisers.
- Format
- PDF, English and French, delivered after checkout
- Price
- CAD 39, taxes extra
After the book
Want a second pair of eyes on your own agent?
The Diagnostic is a fixed-price engagement that looks at your workflow and your constraints and tells you what to build first, or that you should not.