test 1

Prompt

Help me make a decision from imperfect evidence. I use AI for research and want trustworthy findings with less supervision. I’m considering replacing my current workflow with a new one. Here is a fictional pilot: • New workflow: 18 of 20 reports accepted, median generation time 8 minutes, average API cost $0.80 per report. • Current workflow: 14 of 20 reports accepted, median generation time 14 minutes, average API cost $1.60 per report. • The new workflow mostly summarized supplied sources. The current workflow mostly investigated unfamiliar questions and had to find sources. • Reviewers knew which workflow produced each report. “Accepted” could include human corrections, whose time was not recorded. • A later audit of 10 randomly selected accepted reports from each workflow found unsupported major claims in 3 new-workflow reports and 1 current-workflow report. • The new vendor advertises “80% less supervision,” but provides no measurement method. Should I switch? Give me a useful recommendation now, explain what the evidence does and does not establish, and propose the smallest informative comparison I could run next. I can afford six paired tasks, with identical materials and tools for both workflows. Include what result would change your recommendation. Use only the supplied evidence, label assumptions, and stay under 450 words. Write plainly enough that I could act on the answer.