Six threads of inquiry into what it takes to build, govern, and deliver AI systems that enterprises can actually trust — on Google Cloud, at scale, in the real world.
Most architects see the EU AI Act as a compliance burden. I want to argue the opposite: it's the forcing function that finally compels enterprises to build AI systems the right way. Explainability, human oversight, audit trails — these aren't constraints. They're the architecture.
Every regulated enterprise with an AI project is facing this now. Article 14 requirements are not optional from August 2026. If your architecture doesn't satisfy them structurally, you're retrofitting — and that's expensive.
After 15 years inside enterprise Quote-to-Cash systems, I can tell you exactly why every AI POC in this space fails — and what the architecture looks like when it doesn't.
Every enterprise AI project I've seen fail had one thing in common: no architecture. Here's why TOGAF ADM is the scaffold that makes agentic systems trustworthy.
AI in the Business, Data, and Application layers of large enterprises. How to design for scale, not just POC.
Designing product-centric and user-facing AI that closes the loop between action and outcome.
GCP infrastructure, cost optimisation, and the architectural decisions that keep complexity under control at scale.
Managing the AI transformation lifecycle. SAFe delivery, change management, and what actually gets AI to ship.
Conceptual deep-dives into complex system design — agents, ML pipelines, XAI, and the architecture of autonomous systems.
Opinions, provocations, and observations on the state of enterprise technology. The uncomfortable stuff.
No newsletter cadence. Just a note when something worth reading is published.