Escalation Boundary Mapping (Onboarding)

Escalation Boundary Mapping (Onboarding)

1. 01

   Record the client's last 20 real inbound calls and tag each one by outcome

   Sort them into booked, resolved without a human, transferred, abandoned, and callback-requested. The ratio sets the ceiling on what any voice agent can absorb before a retainer is quoted.

2. 02

   Write down the three questions the agent must never answer alone

   Pricing exceptions, clinical or legal advice, and account cancellation are the usual candidates. Each one becomes a hard transfer rule rather than a model judgment call.

3. 03

   Define the transfer destination for every escalation class

   Name the queue, the hours it is staffed, and the fallback when it is closed. A transfer rule without a destination is a dropped call with extra steps.

4. 04

   Set the confidence threshold that triggers a live handoff

   Pick a number and test it against the recorded calls. Too low and the agent transfers routine bookings; too high and it improvises on questions it should have passed on.

5. 05

   Confirm consent and recording disclosure language with the client's counsel before go-live

   Two-party consent states and regulated verticals such as healthcare and legal change the opening line. Trillet's identity-verification and audit-trail design exists for exactly this constraint, and it is worth checking whether the chosen platform records consent at the same point.

6. 06

   Document the labor that stays human after deployment

   List who reviews transcripts, who updates the knowledge base, and how many hours per week that consumes. That number belongs in the retainer scope, not in a footnote.

7. 07

   Run a two-week parallel period where every escalated call is reviewed by a person

   Compare the agent's transfer decisions against what a human would have done. Log each mismatch with the transcript timestamp so the client sees the correction, not just the claim.