AI Agent Protocols for External Stakeholder Interaction

Triggering a Tracing Agent:

• Protocol: If a known potential beneficiary is untraceable through all initial contact methods (phone, email, postal address), the AI agent must flag the case.
• Action: The agent must automatically generate a request for a human administrator to initiate a formal tracing process with an approved tracing agent, as per PSSPF rules. The AI's documentation will record this action and the date it was taken.

Escalation to a Human Claims Officer:

• Protocol: The AI agent is designed to handle standard cases. Any of the following will trigger a human escalation:
  • Conflicting information between documents.
  • A formal dispute lodged by a beneficiary via the provided channels.
  • A request for a discretionary payment waiver.
  • Any legal correspondence received regarding the claim.
• Action: The agent will immediately halt the automated process, notify the designated human claims officer, and provide a full case file with all collected information and the reason for the escalation.

Guidance on Lodging a PFA Complaint:

• Protocol: If a beneficiary expresses dissatisfaction with a decision, the AI agent's primary function is not to defend the decision but to provide clear, actionable guidance on the formal complaint process.
• Action: The agent will provide the contact details for the Pension Funds Adjudicator and a link to their website, along with a brief explanation of a beneficiary's rights under Section 30A of the Pension Funds Act. The agent will then log that this information was provided.