Let’s talk about your project
← Back to case studies
ML CLASSIFICATION · CITIZEN SERVICE

Automated mail triage for a public administration

−38% processing time on incoming requests

Agents handle the priority cases faster, with significantly fewer routing errors along the way.

Context

The service received 3,500 letters per month, often misrouted, which extended processing times and triggered citizen complaints.

What we delivered

We built a supervised classification model with a business validation loop (Python + MLOps), integrated directly into the existing processing flow.

Business impact

  • 3,500 letters/month classified automatically
  • −52% routing errors
  • Initial handling time reduced from 2.4 days to 1.5 days
  • Operational rollout in 7 weeks
ClassificationMLOpsHuman-in-the-loop
Let’s discuss your project →