CASE 01 · CASE ATLAS · LEVEL B WORKSHOP SCENARIO

Hub casting risk analysis

SITUATION

A hub-casting supplier submits a design change. The risk engineer must update the risk analysis for a casting family that already carries three open field observations, under a delivery deadline that leaves little time for a full manual re-derivation.

EVIDENCE

The prior risk analysis, the supplier's change note, the casting drawing revision history and the three field-observation reports, each tagged with casting lot, revision and location.

AI APPROACH

Retrieval surfaces the prior failure modes and their linked field observations against the new change note; language reasoning drafts candidate new or revised failure modes for the engineer to challenge, rather than issuing a finished risk analysis.

KEY CONTROLS

Every drafted failure mode carries its source paragraph and revision. The engineer must challenge, not just accept, each line before it is added to the record.

VALIDATION MEASURES

Measure how many known field-relevant failure modes the draft surfaces versus a manual baseline, and how often the engineer rejects a drafted line as unsupported by the cited source.

HUMAN DECISION

The risk engineer decides which drafted failure modes enter the formal risk analysis and signs the revision; the draft has no standing until accepted.

DISCUSSION QUESTIONS

  1. What in the drafted risk analysis would you refuse to accept without seeing the source paragraph?
  2. How would you detect if the draft quietly missed one of the three known field observations?
  3. What manual baseline would justify running this on the next casting family?

TRY IT LIVE

Run this scenario's approach through Risk Classifier — What level of control does this use case require?

Open Risk Classifier →