Question

A trust-and-safety team needs to review each user-submitted post against five independent policy checks: hate speech, spam, self-harm, personal data exposure, and copyright violation. Each check is a separate, non-dependent evaluation, and the team wants to minimize total latency per post. Which workflow pattern best fits this requirement?

Accepted Answer

Parallelization (sectioning) — run the five checks simultaneously and aggregate the results

Answer

Prompt chaining — feed each check's output as input to the next check in sequence

Answer

Orchestrator-workers — have a model dynamically decide which checks to create at runtime

Answer

Evaluator-optimizer — generate one combined verdict and loop until an evaluator approves it