Whenever you're ready, walk me through how you'd approach this.
How do you design and analyze A/B tests for new AI features?
Describe an end-to-end experimental approach, including randomization, primary and guardrail metrics, sample-size planning, hypothesis testing, confidence intervals, and rollout criteria. Explain how your analysis would address interference, novelty effects, metric skew, multiple comparisons, and safety risks specific to AI features.