The AI data interview

Data interviews still have the same rounds: SQL, a case, product sense, a metrics question, a take-home. What they test is different now. When AI can write the query, the SQL round checks whether you can spot the flaw in one. When AI can produce the analysis, the case checks whether you can frame the question, pick the metric, and say what you would check before you believed the number. That is our read of it. We also read more than 8,000 live US job postings in 2026 to see where AI is already required and where it is barely mentioned. That post is below.

The practice we teach for interviews is the practice we teach for the job. Turn the vague question into a decision. Name the metric and its definition. Walk the trade-offs out loud. Then use AI the way you would at work, to draft, to check, to run the numbers, and keep the judgment yourself. The question that tests this directly, "how do you use AI?", has its own guide with a rubric and three worked answers.