AI Strategy
The AI Investment Frontier
A decision framework for where agentic AI creates durable value, and where to walk away. Value density, the reliability ceiling, and defensibility, scored against an investment frontier.
Knowledge · White Papers
The deeper read.
Long-form analysis from the people who operate agentic AI in production. Each paper takes a single question seriously, the profession forming around AI operations, the economics of metered intelligence, the architecture that holds production together, and the governance a board can sign off on. We argue from what running AI actually requires, not from a roadmap we are selling.
The Library
Six papers on running AI in production.
Strategy, engineering, operations, assurance, data, and cost, each published in full, with no form in front of it.
Agentic AI Engineering
Engineering Agentic Systems That Hold in Production
Why pilots pass and production fails: the compounding-error problem, a five-plane reference architecture, and the controls that make multi-agent systems reliable enough to trust unattended.
Managed AI Operations
Operating AI to a Standard
Production AI is an operation, not a deliverable. A control loop, five operating disciplines, and service-level objectives that keep agents to a standard as models and data drift beneath them.
AI Assurance & Governance
Verifiable AI
AI that acts must be AI you can prove, explain, and defend. An assurance architecture, evaluation, traceability, policy, evidence, mapped to NIST AI RMF and the EU AI Act.
Data Foundations
Data Foundations for Agentic AI
Most agent failures are data failures. Retrieval, freshness, provenance, and the permission boundary an agent must inherit, plus a five-level data-readiness model.
AI Unit Economics
The Unit Economics of Intelligence
When intelligence is metered by the token, reliability is a cost lever: cost per successful outcome, model routing, and the margin discipline that keeps a deployment solvent at scale.
Engage
Begin with a Charter.
A paper frames the question; a Charter answers it for your organization, a fixed-scope diagnostic that maps where agentic AI earns its place and what it would take to run it. Start there, or talk to the people who would operate it.