Pre-Deployment Call Data Intake (Onboarding)

Pre-Deployment Call Data Intake (Onboarding)

1. 01

   Pull 30 days of raw call recordings and transcripts from the client's existing phone system

   Request the full export, not a curated sample. Ask for recordings covering peak hours, after-hours, and weekend traffic so the agent's training set reflects real caller behavior rather than a best-case slice.

2. 02

   Tag every call by intent: booking, qualification, support, spam, or abandoned

   A 500-call sample is enough to establish ratios. If abandoned calls exceed 15 percent of inbound volume, the missed-call follow-up use case is the strongest entry point for the engagement.

3. 03

   Document the current handoff rules between phone staff and the CRM

   Map which fields get written, by whom, and at what point in the call. Platforms like Trillet and ConvoCore write directly into live systems, so any undocumented manual step becomes a gap the agent will inherit.

4. 04

   Measure baseline response gaps: average ring time, hold duration, and callback rate

   These three numbers become the before-state in the client report. Without them, the retainer conversation six months later has no anchor.

5. 05

   Confirm consent and recording requirements for the client's jurisdiction and industry

   Healthcare, legal, and financial services clients carry stricter rules. Verify whether two-party consent applies and whether the client's existing disclosure language covers an AI agent.

6. 06

   Estimate the labor that remains after deployment, not just the labor removed

   List the escalation paths a human still owns: complex complaints, pricing negotiations, and anything the agent cannot resolve in two turns. This figure drives the pricing model more than call volume does.

7. 07

   Price the offer only after the intake data is complete and signed off by the client

   Usage cost per minute varies widely across platforms such as Autocalls, Goodcall, and Vapi. Quoting before the call mix is known produces retainers that lose money on high-volume support traffic.