Reviewing agent performance
Improvement work happens in three places: Call logs (what happened on each call), the Agents view (roll-up and version history), and—when you use them—Follow-ups and Call analysis outputs.
What the static should show: Call details with transcript and recording, or the Call logs table filtered by campaign or agent.
Call logs and call details
- List: On Call logs, find sessions; use your team’s filters and search as implemented on that page.
- Detail: Open a call from Call logs to see Transcript, Play recording / Open recording in new tab (when a recording URL exists), and details such as Voicemail / answering-machine hints when the backend provides them. Review openings, discovery, and how objections landed against what you wrote in Step 4 — Prompt & Goals.
Use exports or bulk actions if your plan includes them (see the Call logs screen for Export / Export history when present).
What the GIF should show: Open a call → scrub audio → jump to a transcript segment.
Agent overview and follow-ups
From Agents, open the agent you care about. On the agent page:
- Config summary shows Model, Language, Voice, Tools count, Follow-ups / Memory toggles, Webhooks (pre-call / post-call), and Knowledge (linked knowledge base or custom).
- If Follow-ups are enabled on the agent, a Follow-ups card can show Conversion and volume (trend/aggregate UI as rendered).
- Linked follow-up rules lists rules attached to the agent; the description points you back to the wizard Step 7 or the follow-up rules page for changes.
- Version history lists snapshots; Restore loads a version as a new draft (per card copy).
- Call (browser) links to a test call with tenant and agent query params.
- Edit agent returns to the wizard to change the prompt, tools, or defaults.
What the video should show: Compare two sessions before/after a prompt edit, then show Version history on the agent.
Call analysis templates (post-call)
Tenant Call analysis prompts are configured on the Call analysis prompts page (LLM provider/model, template text, optional JSON output schema, starter templates). They are for analyzing completed calls, not the live voice script. If your agent attaches analysis in Default tools (Step 7), review those outputs alongside transcripts when tuning extraction or coaching rules.