Transfer Categories and How to Validate Each

Edited

1. Intent-Based Transfers

These are configured outcomes of an intent.

Example:

  • “Request officer” → Transfer to call-taker

These are expected.

How to Validate Intent-Based Transfers (Step-by-Step)

  1. Navigate to Configuration → Intents.

  2. Open the relevant intent.

  3. Confirm:

    • Action is set to Transfer.

    • Destination number is correct.

    • Fallback number is defined.

  4. Review time-based logic for this intent.

  5. Place a controlled test call.

  6. Trigger the intent.

  7. Confirm:

    • Call arrives at correct line.

    • Caller ID behavior is correct.

    • Transfer reason matches expected category in Triage.

Unexpected behavior typically indicates time-based or telephony misconfiguration.


2. Human-Requested Transfers

Triggered when caller explicitly asks to speak to a person.

Behavior:

  • Immediate transfer.

  • Intent flow may be bypassed.

These are normal and healthy.

How to Confirm Human-Requested Transfer Behavior

  1. Place test call.

  2. During conversation, say:
    “I want to speak to a dispatcher.”

  3. Confirm:

    • Immediate transfer occurs.

    • No additional questions asked.

  4. Review Triage:

    • Confirm transfer reason reflects human request.

If system does not transfer immediately, escalate for review.


3. Safety Escalation Transfers

Triggered when:

  • Risk language detected

  • Violence or weapons mentioned

  • Ambiguity creates safety concern

Safety escalation overrides all intent logic.

How to Evaluate Safety Escalations in Review (Step-by-Step)

  1. Open call in Triage.

  2. Review transcript for trigger language.

  3. Open timeline view.

  4. Confirm:

    • Safety override event logged.

    • Intent routing bypassed.

    • Transfer reason aligns with safety trigger.

  5. Determine if escalation aligns with agency policy.

Single safety transfers are expected.
Clusters require pattern review.


4. Jurisdiction-Based Transfers

Triggered when caller location falls outside agency boundary.

May require:

  • Location verification

  • Adjacent PSAP routing configuration

How to Validate Jurisdiction Routing (Step-by-Step)

  1. Confirm jurisdiction boundaries are configured.

  2. Verify adjacent PSAP routing number.

  3. Place controlled test call from out-of-area location.

  4. Confirm:

    • Location correctly identified.

    • Call transfers to appropriate jurisdiction.

    • No routing loop occurs.

  5. Validate caller ID presentation.

Improper routing typically stems from telephony setup, not intent configuration.


5. Guardrail Transfers

Triggered when:

  • Maximum call duration reached

  • LLM timeout exceeded

  • Retry limit exceeded

  • Language instability threshold reached

Guardrails protect system stability.

How to Evaluate Guardrail Transfers (Step-by-Step)

  1. Open timeline in Triage.

  2. Check total call duration.

  3. Look for:

    • Timeout events

    • Retry attempts

    • Duration cap trigger

  4. Confirm transfer reason reflects guardrail type.

  5. Compare against configured thresholds.

Frequent guardrail transfers indicate provider or threshold misalignment.


6. Unknown or Agent Failure Transfers

Triggered by:

  • Provider instability

  • Unexpected system error

  • Timeout recovery failure

These should be rare.

How to Investigate Agent Failure Transfers (Step-by-Step)

  1. Filter calls in Triage by transfer reason = agent_failure.

  2. Identify:

    • Frequency over time

    • Time-of-day patterns

  3. Open timeline for sample calls.

  4. Check:

    • STT errors

    • LLM timeouts

    • Retry failures

  5. Escalate repeated patterns to Admin or Prepared Support.

Isolated failures are expected. Clusters are not.