Skip to main content
Cara ships two analytics dashboards under Settings → Analytics — one focused on Copilot workspaces, one focused on Voice AI. Both dashboards are scoped to your agency and update in near real time.

​
Workspaces

The Workspaces dashboard measures how your team is using Copilot and where Cara is getting stuck.

​
Volume & activity

  • Workspaces created — over time, sliced by preset and by user
  • Active users — who ran at least one workspace in the window
  • Preset mix — which prebuilt and custom presets are pulling weight

​
Preset performance

For every preset (prebuilt or custom), the dashboard reports:
  • Completion rate — Idle vs Waiting vs Working outcomes
  • Median time-to-idle — how long the average run takes end-to-end
  • Follow-up rate — how often users continue the workspace after Cara’s first pass
Use preset performance to spot custom presets that need prompt tuning or Prebuilt Presets that consistently need input from a user.

​
Tool failure breakdown

Cara’s underlying tools (integration reads, code execution, document generation, knowledge searches) are grouped by category. The tool-failure breakdown shows failure counts and rates by tool, so you can:
  • Correlate a failure spike with an integration outage
  • Identify systematically slow tools
  • See when a tool is falling back or retrying

​
Voice AI

The Voice AI dashboard measures call volume and quality across every voice agent your agency operates.
  • Calls handled — over time, sliced by inbound / outbound / agent
  • Airtime — total minutes handled, aligned with the VOICE_MINUTE billing meter (see Usage-based billing)
  • Peak hour distribution — helps size fallback numbers and business hours

​
Disconnection breakdown

Every call ends in a resolution. The disconnection breakdown categorizes each call and reports the counts and percentages:
  • Task completed — Cara handled the caller’s request end-to-end
  • Transferred — Cara routed the caller to a human (warm, cold, or agentic)
  • Voicemail / no answer — Cara detected no human on the receiving side of a transfer
  • Caller hung up — Cara did not complete the task before the caller ended the call
  • Timeout / error — the call ended abnormally

​
Abandonment analysis

For calls that ended in Caller hung up, the abandonment analysis surfaces:
  • Where in the conversation the hang-up occurred (greeting, intake, transfer, wrap-up)
  • Which intents are most likely to abandon
  • Common last utterances so you can identify friction points in Cara’s prompts
Use the abandonment view to iterate on Voice AI configuration and Transfer Directory entries.