AI-Native Engineering Leadership Cohort

The agentic mindset: Context, Reason, Act, Verify, Repeat

Claude Code as a reasoning engine, not a code editor. CLAUDE.md as team knowledge infrastructure. Prompt patterns that work (imperative, exploratory, constrained, verification, decomposition). Token budgets and how they shape work decomposition. Why context engineering matters more than prompt engineering.

Claude Code hands-on: plan mode, skills, hooks, sub-agents

Live demo building a real workflow. Plan mode for non-trivial work. Creating slash commands, skills and hooks. MCP for extending Claude to external systems. The explore-plan-code-commit workflow. Participants follow along on their own machines.

Diagnose / Measure / Scale (introduction)

First look at the Diagnose-Prove-Scale framework. The 7-dimension capability maturity model. Adoption Levels 0-5. The upstream rework loop as the universal constraint. Why you must diagnose before you prescribe. We come back to this in depth in Week 5.

Show & Tell #1

Peer-led. Personal build track demos. Early experiments with Claude Code on your real codebase, what worked, what surprised you.

Architecture for AI-native greenfield

Why architecture decisions are the new bottleneck when execution is cheap. The AI System Architecture Canvas (a 9-section design tool we use on every greenfield). Designing human-AI boundaries explicitly. Model selection and fallback strategies. How bad design decisions compound faster with AI because you build faster.

Live build: napkin to working prototype

End-to-end greenfield using SpecKit. Define the constitution, write the spec, run clarify, plan the architecture, break into tasks, implement phase by phase. Git commit after every command. Review within phases, not just at the end.

Org Level Changes

What changes at the organisational level when engineering goes AI-native. Team sizes shrink from 10+ to 3-5 person squads. Handoffs collapse. Middle management spans widen. The Centre of Excellence (CoE) as an enabling platform rather than a bureaucratic layer. Risk-tiered approval frameworks that let experimentation happen freely while protecting production.

Show & Tell #2

Peer-led. Greenfield demos. What worked, what surprised you, what you would do differently next sprint.

The brownfield challenge: comprehend before you change

Why context extraction is the most important and most time-consuming step on any existing system. BMAD project context generation. ChunkVault for semantic search across legacy codebases. The scout agent pattern for multi-repo systems. Characterisation tests as the safety net before any agent touches legacy code.

Modernisation strategies and the decision framework

Retain / Refactor / Replatform / Replace / Rewrite - when each one is the right call. AI-powered code translation between languages. Incremental migration techniques. Why characterisation tests matter twice as much when an agent is doing the rewrite.

People and Roles change

The people dimension of AI-native transformation. Three new roles emerging (Product Designer, Agentic Experience Engineer, Product Engineer) and the roles that compress. The change curve - Denial, Resistance, Exploration, Commitment - and what intervention works at each stage. Why the no-involuntary-departures commitment is non-negotiable. The Knowledge Paradox - AI amplifies expertise for seniors but creates a confidence trap for juniors.

Show & Tell #3

Peer-led. Brownfield experiments. Modernisation stories. What broke first when you pointed an agent at production-adjacent work.

Organisational diagnosis and the business case

The 7-dimension capability maturity model. Adoption Levels 0-5. The upstream rework loop as the universal constraint. How to build a credible business case from a diagnosis - the language and structure that lands with CFOs and boards.

Running a diagnosis in practice

Live walkthrough of a real diagnosis. The interview pattern, the artefact pull, the metrics that matter and the metrics to ignore. Three-layer metrics: team-level (cycle time, WIP, throughput), leadership-level (DORA, rework rate, value-added ratio), board-level (capacity created, cost per feature, strategic capability). Outcome metrics versus vanity metrics.

Diagnose -> Scale: Client case

End-to-end walkthrough of a real client transformation, from initial diagnosis to scaled adoption. What we measured, what we changed, what worked and what did not. The capacity-based ROI model for the CFO conversation.

Show & Tell #4

Peer-led. Diagnosis findings from your own organisation. Stress-test each other's intervention roadmaps. Where would you start tomorrow morning?

Building the agentic platform - Champions and CoE infrastructure

The agentic platform as enabling infrastructure - shared agents, skills marketplace, governance baked in. The Champions network and the CoE pattern (centralised, hub-and-spoke, embedded) and when each works. How to fund and staff the platform without standing up a new bureaucracy.

AI-native beyond engineering

Cross-functional applications: product (AI-assisted discovery and validation), design (code-first review replacing months of Figma cycles), HR (recruiting, onboarding, policies), marketing (content, campaigns, analysis), operations (incident response, documentation, reporting). The pattern for picking which function moves second - and why the wrong choice stalls everything.

KPIs and impact tracking

How to measure transformation impact honestly and present it to different audiences. Outcome metrics vs vanity metrics - why tracking tool adoption tells you nothing about whether the transformation worked. Rework rate as the truth test. The capacity-based ROI model for the CFO conversation. How to run the Day 90 readout with honest numbers, including the things that did not work.

Personal Build Showcase + Graduation

Peer-led. Extended presentations of what you built and what the metrics moved. Certificate handover. Transformation toolkit and alumni network onboarding. The room you will keep coming back to.