The Three-Tier Architecture for AI in the PMO
The operating model behind multi-agent AI for the PMO: a layered architecture where data reconciliation, practitioner sense-making, and director-level judgment each get their own tier.
Architecture, decisions, and field notes on AI in the project management office.
I run an Enterprise Project Management Office and have spent the last year building a multi-tier AI architecture that gives PMs near-realtime awareness of project and portfolio data, surfaces signal across the organization, and supports the strategic decisions PMOs are usually too far from the data to make well.
This site holds the operating model, the architecture, the case work, and the field notes. The vendors selling AI to PMOs right now are mostly selling the wrong thing. The right thing is buildable, but it has to be built — not bought.
The operating model behind multi-agent AI for the PMO: a layered architecture where data reconciliation, practitioner sense-making, and director-level judgment each get their own tier.
AI does not replace project management. It moves the discipline up the cognitive stack — handling reconciliation and aggregation so practitioners can operate where the value actually lives.
Most AI products being sold to PMO leaders right now are shaped to clear procurement and demo well, not to produce operational value. The gap between the pitch and the reality is going to widen.