Radical Geek field guide
Radical Geek's Guide to the Agentic Engineering Operating Model
A practical operating model for human accountability, governed work classes, workflow bundles, evidence, approvals and continuous learning.
Why this guide
Agentic engineering becomes a team capability when people can run it through clear roles, approved workflows, evidence requirements, decision rights and recovery paths. Individual tool use can help a developer; dependable delivery requires an operating contract.
This guide defines how humans and agents share software work while people retain product judgement, system design, production risk, customer impact and final accountability.
Use it to shape the daily team system around the wider Radical Geek guide series: governance, repository readiness, security, patterns, model lanes, evaluation, observability and business impact.
Make ownership, evidence, decision rights and learning loops part of normal engineering delivery.
What you will leave with
A guide built to be used.
- 01
Define accountable human and agent roles with explicit authority, evidence and decision rights.
- 02
Establish a small set of approved workflow bundles for repeatable classes of engineering work.
- 03
Build planning, assignment, review, merge and release rituals around inspectable evidence.
- 04
Use incident, defect and learning loops to improve prompts, tests, routes, permissions and work-class policy.
- 05
Align enterprise architecture, platform engineering, skills and management metrics around agentic delivery.
Inside the guide
The working ground it covers.
- Roles and responsibilities
- Workflow bundles for teams
- Planning, assignment and review rituals
- Human approval, merge policy and release governance
- Incident, defect and learning loops
- Skills, platform ownership and metrics
- A practical first 30 days plan