Building agent judgement calls with Microsoft Decision-1 scores and thresholds
marlene_zw · x · 2026-11-06
A Microsoft engineer shares how she gave her Teams/Outlook agent "Mona" judgement: Microsoft Decision-1 provides fast 0-1 scores for yes/no questions, and a few lines of Python turn them into decisions.
- Three scenarios: whether to jump into an untagged conversation (noise ≥0.6 → pass); whether a drafted channel post is too sensitive (money/HR/secrets flags → hold, fails safe); whether a finished exchange belongs in her personal feed (signal ≥0.6 → show, routine ≥0.7 → hide).
- Architecture: a Microsoft Foundry-hosted agent with gpt-6-luna for reasoning and Decision-1 for fast scoring; when the model is down, the agent defaults to staying quiet.
- She also highlights WorkIQ as a memory store that scopes memories by contact and channel — the most valuable tool for agents in the Microsoft ecosystem.
Related event: Microsoft Engineer Shares How She Taught Agents Judgment(2 posts)→
More from coding & agent
- The AI discount trap: why demanding cheaper AI-assisted development can cost you more — luisdans · 2026-11-07
- Founder wires an agent to check his product's ranking every 10 minutes and ping Discord — pramodk73 · 2026-11-07
- Two engineers argue whether prompting still matters in the agent era, with a pipeline-fix prompt example — teodorio · 2026-11-07
- Hermes-Agent hacks Echo Show 5: screen is just a web page you can customize — alexcovo_eth · 2026-11-07
- 5 vibe coding component libraries worth pasting into Claude or Cursor — thisiskp_ · 2026-11-07
- Design engineer automates his own $56K/yr drafting job with an AI drawing tool — luisdans · 2026-11-07