Scenario
A HAZOP or safety review is only as good as its preparation. Teams often walk in without a structured view of the reactive hazards, incompatibilities, and thermal risks in the process. A first-pass safety triage gives the review an evidence-based starting point.
Worked example. A new solvent swap introduces a potential peroxide-former and an incompatibility with an existing reagent.
Agent workflow
- Read the process: materials, quantities, and conditions.
- Process safety triage — screen for reactive-hazard classes, known incompatibilities, and thermal-stability concerns.
- Compile a prioritized hazard briefing for the review team.
Demo output
A hazard briefing that lists each material's reactive classification, flagged incompatibilities, and thermal concerns, ranked by severity. The magic moment: the triage catches an oxidizer/fuel incompatibility that would otherwise surface only during the walkthrough.
Deterministic vs LLM
- Deterministic — the hazard-class screening and thermal-stability rules.
- LLM — the safety-review briefing narrative.
Every flag traces to a rule or reference. The agent reads and routes; the engine decides.
Limits
Preparation-level screening, not a substitute for a formal HAZOP, LOPA, or calorimetric testing. It sharpens the review; qualified assessors sign off.
Skills orchestrated by this use case
This scenario routes an AI agent through the following curated skills.
Tools it reaches for
- Reactive-hazard and incompatibility screening
- Thermal-stability estimation
Browse the full deterministic layer in the tool browser.