Rail and manufacturing
Months of compliance reading, done in minutes
Rail systems manufacturerSafety-critical certificationGermany
The problem
A German manufacturer of safety-critical rail systems ran certification the way it had been run for decades. Senior engineers cross-referencing product specifications against regulation and against every previous certification, by hand. Each pass took weeks to months.
What shipped
- An agentic pipeline that reads the specification, the applicable regulation and the certification history together
- A supervisor and worker architecture that keeps each agent scoped to one job
- Every conclusion tied back to its source documents, an answer with its evidence attached
- An engineer review and approval step on every output
Problem-to-production blueprint
1DiagnosisMapped the manual certification pass: which documents were cross-referenced against which, where the weeks went, and what a defensible conclusion has to show.
2Architecture & PoCA supervisor and worker agent design, each agent scoped to one job, proven on real specifications, regulation and certification history.
3Production engineeringEvery conclusion traced back to its source documents, so the output is an answer with its evidence attached, ready for audit.
4StewardshipEngineers review and approve each output; senior time returns from document work to engineering.
Outcomes
MinutesCompliance cross-reference
Full traceEvidence path per conclusion
Senior hoursEngineering time returned
The wider point
What this means for you
If a regulator reads your outputs, AI is an evidence problem before it is a technology problem, and getting it wrong is a liability rather than a bug. The question is whether your data, governance and controls can support AI a regulator will accept, before you build it.
Same problem, different regulator
Bring us the compliance work a regulator will read.
We diagnose before we build, and one owner takes it from first decision to production.