Category Definition
AI/CD
Agentic Integration / Continuous Delivery
The governance framework for agent-driven software delivery. AI/CD moves integration from human-driven to agentic while keeping human delivery gates.
Coined by Torben Anderson, February 2026. Built in the open.
The shift
CI/CD assumed humans write code and machines build, test, and deploy. That assumption is broken.
AI agents (Cursor, Copilot, Claude Code, Devin) are writing code, running tests, proposing reviews. But the governance model hasn't caught up. GitHub calls their approach "Continuous AI." PwC calls it "Agentic SDLC." Neither provides a governance framework for how enterprises manage the transition safely.
AI/CD is that framework.
The evolution of delivery
Continuous Integration / Continuous Delivery
Humans implement. Automation builds, tests, deploys.
Who integrates? Humans
Who gates delivery? Humans
Agentic Integration / Continuous Delivery
Agents assist implementation, tests, reviews. Humans retain gate and release control.
Who integrates? Agents
Who gates delivery? Humans
Agentic Integration / Autonomous Delivery
Agents execute under policy. Humans set strategy and handle exceptions.
Who integrates? Agents
Who gates delivery? Agents (policy)
Note: "AI" in AI/CD stands for Agentic Integration — the integration phase itself is now agentic. Not "Artificial Intelligence." This is a deliberate reframing.
The governance model
Gate-driven lifecycle
AI/CD is built on a gate-driven lifecycle: 6 steps, 5 gates, production feedback loop. Every gate is an audit point — steering probabilistic AI output back to deterministic quality standards. Evidence-first: every gate transition links to an immutable artifact (commit, PR, issue comment, workflow run, or release tag). Platform-agnostic: works on GitHub, GitLab, Azure DevOps, Jira + any SCM combo.
- S1: Strategy and Intake → G1: Load and Scope Ready
- S2: Packet and Spec Preparation → G2: Spec and Contract Aligned
- S3: Implementation and Tests → G3: Test and Behavior Verified
- S4: Workflow and AI Harness Controls → G4: Operational Guardrails Passed
- S5: Done Gate Evidence Review → G5: Closeout Criteria Met
- S6: Merge Hygiene and Release → O1: Production Feedback Loop (cycles back to S1)
Gates act as stage gates and audit points — steering probabilistic AI decisions and output back to quality management best practice defined by the organisation.
Reference implementation
Icebox CLI is the open-source reference implementation of AI/CD governance.
Real artifacts: actual PRs with gate evidence, issues with comment trails, milestone tracking. This isn't a whitepaper — it's a working system. Fork it, adapt it, implement it.
Who this is for
- Engineering leaders evaluating how to bring AI agents into delivery safely
- CTOs and VPs of Engineering already using Cursor, Copilot, Claude Code — who need governance around agent output
- Regulated industries (finance, healthcare, government) where audit trails are non-negotiable
- Startups scaling from founder-led development to agent-assisted teams
- Fractional CTOs and consultants who need a repeatable framework for client engagements
Work with Rewired
Implementing AI/CD at your company? Rewired helps engineering teams adopt agentic delivery with the governance enterprises require.
Get Clear session — $797
90-min diagnostic. Assess your current delivery pipeline and map the transition to AI/CD.
AI/CD Implementation
Hands-on engagement to stand up AI/CD governance in your pipeline. Pricing depends on scope. For startups, we also offer a productized option: $7,500 setup + $2,500/mo.
Built by Torben Anderson — 40 years shipping production systems. This isn't theory. I run AI/CD across every client engagement. You talk to the person who builds it.
Related reading
FAQ
What does AI/CD stand for?
AI/CD stands for Agentic Integration / Continuous Delivery. It's the next evolution of CI/CD where AI agents handle integration (writing code, running tests, proposing reviews) while humans retain gate and release control. Note: "AI" here means "Agentic Integration" — the integration phase itself is now agentic. Not "Artificial Intelligence."
What's the difference between AI/CD and AI/AD?
AI/CD keeps humans in the delivery loop — agents assist but humans approve. AI/AD (Agentic Integration / Autonomous Delivery) is the next phase where agents execute under policy guardrails and humans set strategy and handle exceptions.
Who coined AI/CD?
AI/CD and AI/AD terminology was introduced by Torben Anderson of Rewired Consulting in February 2026.
Is AI/CD open source?
The reference implementation of AI/CD governance is Icebox CLI, an open-source project. The process framework is available at github.com/torbenanderson/icebox-cli.
How is AI/CD different from GitHub's Continuous AI?
GitHub's Continuous AI adds agentic workflows to existing CI pipelines. AI/CD is a broader governance framework that defines how enterprises transition from human-driven to agent-driven delivery across the entire lifecycle, with explicit gates, audit trails, and compliance evidence.