Deep dives, no email gate.
Long-form writing on the architecture and measurement problems that decide whether an AI programme delivers. Download them directly — we do not gate our best thinking behind a form.
The AI ROI Playbook
Five dimensions to measure, three traps to avoid, and the one question we ask every prospective client before scoping an engagement.
38 pages · PDF
Everything else we have published.
RAG architecture deep dive
The reference architecture we ship for citation-grounded enterprise RAG, with diagrams, tradeoffs, and the failure modes each choice accepts.
- RAG
- Architecture
- Retrieval
Reliability patterns for AI agents
Deterministic checkpoints, replayable traces, and the boring infrastructure that makes autonomous agents trustworthy enough to authorize.
- Agents
- Reliability
- Governance
Model risk governance for LLM systems
Mapping SR 11-7 expectations onto generative systems: validation evidence, monitoring plans, and the documentation your MRM function will ask for.
- Governance
- Compliance
- Finance
We will walk you through any of these.
Thirty minutes with the person who wrote it, applied to your situation rather than the general case.