Intelligence

AI layer

Inline scoring on every meaningful API response. Stable schemas, model versioning, and a full audit trail per decision.

Pierflow is AI-native because health data is too complex for static rules at scale. Every score is returned alongside the data — no separate model calls to orchestrate.

fraud_score

0–100
Returned on enrollments, claims, and payments. Higher = higher risk. Tunable thresholds available per partner.

identity_confidence

0–1
Confidence on the verified identity of a member. Combines BVN, NIN, biometric, and historical signal.

lapse_risk_score

0–1
Risk that a policy will lapse during the next collection cycle. Designed for proactive retention.

value_score

0–100
Plan-quality score balancing benefit breadth, network depth, and pricing efficiency.

eligibility_confidence

0–1
Confidence on a verification result given the policy and provider context.

Auditability#

Every score is logged with the model version, input hash, and decision time. Full traces are queryable from the operations dashboard.

Why this matters

AI in healthcare should work invisibly — improving decisions and catching errors — without ever replacing human judgment. Pierflow scores are explainable signals, not black-box verdicts.

Model versioning#

Each score carries model_version in long-form responses (e.g. fraud_v3.2). Old versions remain queryable for at least 12 months so audits can reproduce historical decisions.