Most organizations start AI in one place: the code editor. But writing code is only one step of delivering software. If everything around it stays the same, faster coding mostly creates longer queues.
The lifecycle, stage by stage
| Stage | AI uses | Watch out |
|---|---|---|
| Plan | Summarize requests, draft specs and tickets, estimate | Specs that sound complete but aren't |
| Code | Completion, chat, coding agents | Context gaps, insecure patterns |
| Test | Generate and extend tests, find edge cases | Tests that assert the wrong behaviour |
| Review | First-pass review, suggested fixes for comments | Reviewers rubber-stamping |
| Release | Release notes, change-risk scoring | False confidence |
| Operate | Alert triage, root-cause suggestions, runbooks | Wrong root causes under pressure |
| Maintain | Migrations, upgrades, documentation | Large changes that are hard to verify |
What the research says about the system
DORA's 2024 report modelled what happens as AI adoption rises. A 25% increase in AI adoption was associated with an estimated 7.5% increase in documentation quality, 3.4% in code quality and 3.1% in code review speed — but also an estimated 1.5% reduction in delivery throughput and 7.2% reduction in delivery stability. DORA's hypothesis: AI makes it easy to produce larger changes, and large batches have always been riskier.
DORA's 2025 research described AI as an amplifier of an organization's existing strengths and weaknesses, and listed seven capabilities that amplify its benefits, including working in small batches, strong version control practices and quality internal platforms.
The CTO's move
Map where time actually goes from idea to production. Put AI where the constraint is, and fix the constraint around it — review capacity, test automation, CI speed — before pushing more code into it.