At Sable Microsystems, an engineering change to an RF power amplifier board used to mean three weeks of chasing. An engineer would open the ECR in ENOVIA 3DEXPERIENCE, then spend days walking the BOM by hand, emailing five departments, collating replies in a spreadsheet, and rebuilding a risk table from memory. This is how their principal hardware engineer, Dana Whitfield, ran the same change end to end in 2.5 business days using Veroli as the front door and VeroCortex agents doing the work underneath.
The problem wasn't the analysis. It was the assembly of it.
Sable's change board never doubted the engineering judgment. What slowed them down was everything around it: finding which items a change touches, knowing who owns them, getting scoped answers instead of "looks fine to me," and proving afterwards that the record in PLM matches what was actually decided. The EC Impact Analysis WorkStream turns that assembly work into six governed stages.
Day 1 · 08:40
Stage 1. The ECR arrives and scopes itself
Dana doesn't fill in a form. She states the change in plain language. The Intake Agent reads ECR-4471 straight out of 3DEXPERIENCE, pre-populates the intake record, and proposes the two fields the ECR never carried: target effectivity and regulatory class, inferred from the last three thermal changes on the same platform.
Day 1 · 08:52
Stage 2. Twelve minutes to a context dossier
The Context Agent walks the BOM instead of Dana. Eleven levels down, it returns 34 affected items, 6 open work orders, 2 supplier agreements, and 4 qualification records, each one resolved to an owning function. Two items fall outside Dana's division, so they route separately rather than dying in a group inbox.
Day 1 · 09:15
Stage 3. Five scoped questionnaires, not one blast email
This is where the old process leaked. A single "please review" email to a distribution list produces vague replies and no accountability. The Stakeholder Agent instead drafts one questionnaire per owner, each asking only about the items that person actually owns, with a named due date. Replies land back in the workstream, and the agent chases anything still open after 24 hours.
Day 2 · 14:20
Stage 4. The risk register writes itself, with scores
All five responses were back inside 27 hours. The Analysis Agent turns them into a scored register: severity, occurrence, and detection multiplied into an RPN, with anything above Sable's threshold of 100 escalated and given a named owner and a mitigation. Every line cites the response or record it came from.
Day 3 · 10:05
Stage 5. Sign-off on evidence, not on trust
The change board doesn't receive a PDF summary and a request for goodwill. Each signer sees the claims, the RPN math, and a link from every claim back to the source record in the Execution Ledger. Three signatures landed in seven minutes.
Day 3 · 10:11
Stage 6. Write-back, then proof that it wrote back
The last mile is the one most automation skips. The PLM Agent moves ECR-4471 to Approved in 3DEXPERIENCE, attaches the impact analysis, and stamps the change note on all 34 affected items. Then it re-reads ENOVIA and reconciles the result field by field, so "the system was updated" is a verified claim rather than an assumption.
What changed
Sable didn't replace ENOVIA, and they didn't ask engineers to work differently. They moved the coordination out of inboxes and spreadsheets and into a governed WorkStream that runs on top of the PLM system they already had.
- 19 days to 2.5 days. Median ECR impact cycle time across the first quarter of use.
- Zero manual BOM traversal. 11 levels and 34 affected items resolved by the Context Agent.
- 100% stakeholder response inside 27 hours. Scoped questionnaires beat broadcast emails.
- Verified write-back. Every PLM update reconciled field by field, with an immutable ledger entry behind it.
The engineering judgment stayed exactly where it belongs, with the engineers. What Sable automated was the three weeks of work that used to sit around it.
