A retail operations team builds an automation that summarizes daily order issues from GitHub issues into a manager briefing. It must be concise, follow a fixed template, and avoid creative embellishment. Stakeholders complain drafts are rambling and inconsistent. Which prompt and model configuration for the automation best improves consistency?
Select an answer to reveal the explanation.
Short Explanation
You'll get tighter briefings when you pair a precise model with a locked template, length limits, and only the facts at hand. That's like swapping a blank page for a fill-form sheet, so your template keeps every draft neat and on track.
Full Explanation
The correct configuration succeeds because it aligns every control to the stated failure mode: a fixed template enforces identical structure across days, explicit length limits force conciseness and cure rambling, factual grounding to the GitHub issues prevents invention, and a precise capable model executed deterministically delivers repeatable adherence that weaker or more stochastic setups cannot guarantee. The lightweight economical alternative looks attractive for routine daily volume and keeps good prompt hygiene, yet it fails on model capability since smaller efficiency-tuned models more often drift from templates and grounding under real-world variation, which reintroduces the inconsistency stakeholders reported. The high-temperature variant also keeps structure and limits but fails on determinism because raising sampling randomness to improve fluency directly increases phrasing variance and hallucination risk, violating the avoid embellishment and improve consistency goals. The extended-length variant keeps template, grounding, and capable execution but fails on conciseness because permitting extra length for completeness removes the cap that forces prioritization, so briefings become thorough yet rambling again. For exams, beware distractors that are best practices in other contexts like cost optimization, livelier tone, or comprehensive reporting but violate one stem constraint, so match every adjective in the stem before selecting. As a practical check, pilot the briefing on a week of labeled issues with low temperature, strict max tokens, schema validation for template fields, and source-citation review to confirm brevity, stability, and faithfulness before rollout.