Define the proposed intervention
Set the account action, customer and organisational consequences and confidence required.
HCIIF IN PRACTICE · FINANCIAL SERVICES
A high-risk alert.
How strong is its basis?
An AI-assisted fraud alert recommends restricting a customer account, but the underlying sources differ in date, purpose and verification.
The information
challenge.
The score is high, but a score is not proof.
The alert combines transaction activity, customer records and an external indicator. One record is current, another has not been recently verified, and the external indicator has limited context.
The organisation must respond proportionately while considering fraud exposure, customer impact and the evidence available.
Trace the alert and its inputs
Use ECAM to assess provenance, currency, verification, model contribution and any material transformation.
Separate indicator from conclusion
At the point of use, show the Confidence Gap and limitations before a score is treated as evidence of wrongdoing.
Keep the basis current
Use the IIP to preserve the confidence basis and trigger reassessment when evidence, customer context or the model changes.
ILLUSTRATIVE INFORMATION INTEGRITY PASSPORT
What confidence is justified?
- Confidence basis
- A current transaction pattern is supported; other inputs have verification and context limitations.
- Justified reliance
- Sufficient for proportionate review, but not by itself proof of wrongdoing.
- Accountability
- The authorised decision-maker retains responsibility for intervention and customer impact.
- Confidence Drift and reassessment
- New verification, customer evidence, model review or a change in transaction context.
HCIIF difference
HCIIF tests whether the alert justifies this intervention, keeping its evidence and limitations visible rather than treating an automated score as a definitive conclusion.
This example illustrates information-integrity management and does not determine a regulatory or customer decision.