NHSAIMachine learningAssurance/Risk

Fraud & anomaly detection

Anomaly detection capability to flag high-risk payment patterns and generate investigation-ready outputs.

At a glance

£2.2M
High-risk transactions flagged
Outputs
Referral-ready investigation packs

Challenge

Large invoice and payment datasets contained patterns that were difficult to detect manually, requiring scalable anomaly detection and triage.

Approach

  • Built anomaly detection using machine learning + LLM-assisted pattern interpretation
  • Implemented triage workflow and escalation outputs
  • Created investigation packs with auditable rationale

Data sources

  • Invoice and payment datasets
  • Supplier master data
  • Approval and exception logs

What was delivered

  • Risk scoring model
  • Triage and escalation workflow
  • Investigation pack outputs

Impact

£2.2M high-risk transactions flagged and referred for investigation
Improved prioritisation of investigation workload