In 30 seconds
- Story points blend effort and uncertainty; AI shrinks effort, not uncertainty.
- Velocity can jump without more value reaching users.
- Shift toward throughput, lead time and outcomes.
Why estimates drift
Story points mix effort, complexity and uncertainty. AI cuts typing effort sharply but leaves uncertainty about requirements and integration. Old reference stories no longer mean what they did, so velocity history breaks.
What to measure instead
| Instead of | Try |
|---|---|
| Points per sprint | Throughput of small items |
| Coding effort | Lead time, including review |
| Output | Outcomes and escaped defects |
Forecasting honestly
- Re-baseline after adopting AI tools; old history misleads
- Size by uncertainty and risk, not typing effort
- Forecast with ranges from recent throughput; see forecasting
Myth vs reality
Myth · Higher velocity proves AI is working.
Reality · It may only show inflated output.
Myth · Estimation is dead.
Reality · Forecasting still matters; the inputs change.
Myth · AI makes everything equally quick.
Reality · Unclear or risky work stays slow.
FAQ
Should we stop using story points?
Consider it if items are small and similar; throughput may be simpler.
How long to re-baseline?
A few sprints of new data.
What about Planning Poker?
Still useful for surfacing uncertainty, even if numbers matter less.