The view is organized by initial handler
Agent vs Human groups calls by who initially answered the customer interaction. It shows call counts, booked calls, booking rate and average duration where the relevant data is covered. A call that starts with AI and later transfers to a person stays AI-first.
That rule prevents a single customer interaction from being counted twice. The human handoff is still meaningful; it is reflected separately as AI-assisted human activity and in the call's transfer state.
Read N/A as unavailable, not zero performance
The current Voice-origin reporting does not provide comprehensive human-first coverage. Human Answered or a human performance comparison can therefore show N/A and a coverage explanation. It does not mean your employees answered no calls or produced no bookings.
Do not calculate “AI beat the office” from an available AI figure and an unavailable human figure. Use Call Spark or the appropriate phone-system report for staff calls outside the Voice dataset, and reconcile the scopes before comparing them.
Match the populations before comparing rates
Use the same time zone, reporting period, lines and service types. An AI backup agent may receive only calls staff could not answer, while a daytime human team handles a different mix. After-hours callers, existing-customer questions and new-service leads also have different opportunities to book.
The displayed booking rate uses all calls in each selected handler group. It is not automatically adjusted for qualified leads, staffing, ad source or service availability. A source or outcome drill-down changes the population again, so check the filters before comparing screenshots.
Separate ownership from business action
Who first answered, who later took over, and who performed a confirmed booking are three different facts. AI-attributed bookings require the appropriate captured receipt; unknown attribution is not proof of a human booking.
Review an AI-to-human call as a shared workflow. Did the agent gather useful details? Did the customer reach the right person? Was the next step completed? A useful escalation is not automatically an AI failure, just as a short call is not automatically good service.
Use the view for coaching
- Identify a specific pattern rather than a single outlier.
- Open calls from that group and review the concern, conversation and saved result.
- Classify the cause: knowledge, service scope, routing, CRM availability, audio or human process.
- Change the relevant configuration or workflow.
- Compare another similar period and inspect representative calls again.
Keep a baseline and record routing changes. Do not set universal performance targets from a small sample or a marketing screenshot. Establish your own expectations from adequately covered, comparable business data.