Weekly analytics digest
Product managers drown in dashboards, so you're asked to build an agent that writes a weekly digest of your product analytics — what moved, what broke, what's worth a look. Design the workflow: where does it pull numbers from, what is it allowed to claim, and who reviews it before it lands in inboxes? How would you check that it's not inventing trends? Finish by pitching it to the head of product in a couple of sentences.
select_digest_claims(names: list[str], current: list[float], baseline_mean: list[float], baseline_std: list[float], band_sigma: float) → list[str][["signups","crashes","revenue"],[1200,45,98000],[1000,50,100000],[50,10,3000],2]out["signups up"][["latency_p95","churn"],[180,0.02],[200,0.05],[5,0.01],2]out["latency_p95 down","churn down"][["signups","revenue"],[1010,99500],[1000,100000],[50,3000],2]out[]Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.