Case study · Codesa S.A.

Biometric identification in production: less fraud, validation in seconds

Client: Codesa S.A. · Facial recognition at service points

Industry: transactional / remittances · Live across the service network

Context

Codesa runs service points where customer identity validation is critical to reduce fraud and speed up operations.

They needed reliable facial recognition wired into the real service-point flow — not a standalone PoC.

Problem

  • Fraud and unverified identity at the counter / service points.
  • Slow manual validation (~5 minutes) that hurt the experience.
  • Latency, security and rollout constraints across dozens of locations.

Approach

  1. Discovery of process, KPIs (fraud, validation time) and existing integrations.
  2. Architecture: facial recognition with AI + Spring Boot backend + Angular frontend.
  3. Pilot at selected service points with precision and latency metrics.
  4. Hardening and rollout to 50+ live service points.

Technical solution

  • Components: facial recognition pipeline, validation API, operator UI in Angular.
  • Integrations: Spring Boot backend connected to the service-point operational flow.
  • Evaluation: match precision, response time and reduction in fraud attempts.

Results

MetricBeforeAfter
Fraud attemptsOperational baseline85%
Validation time~5 min30s
Service pointsLimited pilot50+

Learnings

  • Operator latency and flow matter as much as the model: AI only helps if it fits real operations.
  • A measurable pilot (fraud / time) unlocks rollout; without a clear KPI the project stalls as a demo.