Six frameworks. One methodology behind the work.
Six frameworks, applied across government infrastructure, enterprise AI deployments, and multi-jurisdiction governance programmes. Each is published here as a working reference, not a marketing summary: versioned, with a reference diagram, a suggested citation and a PDF brief. Open one to read the full methodology, or see how they fit together in my book AI at Scale: From Pilot to Production.
Listen to this overview
How to Prevent Enterprise AI Disasters
Six frameworks, six shapes, each the wireframe of the framework it stands for.
01Five-Dimension AI Readiness Assessment
Five operational dimensions, scored 1–4, that predict AI deployment failure before any budget is committed. A single low score blocks the initiative regardless of the others.
Read the framework →
02TRACE Framework
Five pre-deployment checks for whether a task is structurally appropriate to hand to an AI agent, before architecture or vendor decisions are made.
Read the framework →
03Four-Question AI Governance Baseline
Four diagnostic questions that expose structural governance gaps before deployment. The same four questions regardless of scale.
Read the framework →
04Return on Employee (RoE) Framework
Measures AI value through the increase in productive capacity per person, replacing headcount reduction as the default (and misleading) business-case metric.
Read the framework →
05Measurement-Before-Prediction Architecture
A five-step sequence for designing the measurement schema before data collection begins, so AI predictions don’t require months of recalibration after go-live.
Read the framework →
06Governed Cognitive Layer (Layer 8)
The layer enterprise AI needs above OSI’s seven: a semantic router, a reasoning loop and a deterministic guardrail check that turn probabilistic reasoning into governed, auditable action on production systems.
Read the framework →Ways to work together.
Every framework on this page started as a mess: a project going sideways, a client asking a question I couldn't answer cleanly, a gap between what the slide promised and what the system did. I wrote these down so I'd stop solving the same problem twice. If you're building something and the framework doesn't quite fit your situation, change it. That's what I do too. The judgment behind the method matters more than the method.
