The CDI cheat sheet

What the cheat sheet covers

This is version 0.1 of the Cognitive Delivery Index one-page reference, and it is superseded by the CDI Companion Guide. It describes the Index as a six-dimension assessment of an organisation’s readiness to deliver software in an age where humans and agents share the keyboard, set against a spectrum from Dev 1.0, Solo Author, where agents are absent or invisible to the workflow, through Pair Coder and Delegated Tasks, to Dev 4.0, Orchestrated Agents, where humans set intent and constraints and agents plan, execute and report against them.

The six dimensions in this version are Delivery Maturity, Governance Posture, Accountability Clarity, Review Integrity, Adoption Scale and Outcome Trust. Each carries a description of what low and high look like, and a diagnostic question to put to the team.

Key points

• Three of the diagnostic questions: If an agent acted badly today, would we notice within minutes, weeks, or never? When something ships and breaks, can a named human still defend the decision? If the agent disappeared tomorrow, would the team still own what it built?
• Each dimension is scored independently on a five-point scale: 0 Absent, 1 Ad-hoc, 2 Defined, 3 Managed, 4 Systemic. The composite gives a single read on readiness, and the shape of the profile tells you where to invest.
• Red flags signal low scores being masked by surface productivity gains: “we don’t track that yet” said about agent activity, reviewers approving PRs they could not have written themselves, productivity claims without paired trust metrics, and velocity rising while incident rates rise alongside.
• An assessment is a 60-minute structured self-assessment with a cross-functional group: delivery, security, architecture and ops, not just engineers.
• Everyone scores each dimension alone before any discussion. The disagreements are more useful than the averages.
• Act on the lowest two dimensions. The highest does not need investment, and trends matter more than scores.