In 30 seconds
- Most AI anti-patterns repeat old agile anti-patterns, faster.
- Watch for output-over-outcome thinking.
- Fix habits, not just tools.
Common anti-patterns
| Anti-pattern | Instead |
|---|---|
| Measuring AI success by lines of code | Measure outcomes and lead time |
| Rubber-stamp review of AI code | Real human-in-the-loop review |
| Raising velocity targets after adopting AI | Re-baseline; see estimation with AI |
| Tool mandates without team input | Teams experiment and share results |
| Skipping discovery because building is cheap | Keep discovery with real users |
Warning signs
- Rising change failure rate
- Growing review queues
- Nobody can explain recent code
- More features, flat outcomes
Fixing them
Treat AI adoption like any other change: small experiments, visible results, and retrospectives. See also agile anti-patterns.
Myth vs reality
Myth · Faster output means success.
Reality · Only outcomes prove value.
Myth · Resistance means people need more training.
Reality · It may signal real quality concerns.
Myth · Anti-patterns are a tool problem.
Reality · They're mostly habit and incentive problems.
FAQ
What's the most common anti-pattern?
Measuring output instead of outcomes.
How do we spot rubber-stamping?
Very fast approvals on large AI-generated changes.
Who should lead fixes?
The team, supported by leaders who change incentives.