Set Up Post-Call Tools and Analysis
A finished call is full of answers your team needs — intent, budget, location, the agreed next step — and none of it is useful while it's locked in audio. Post-conversation tools run after the call ends and turn it into usable CRM context: a short summary a follow-up owner can read before calling back, extracted fields, outcome labels, reschedule requests, and structured analysis for reports.
Entry pointAutomation → AI Assistants → Post-Conversation Tools
What Post-Conversation Tools Do
| Tool or feature | Use it for |
|---|---|
| Summary | Create a short human-readable call summary. |
| Extract Variable | Capture one specific value such as budget, location, or preferred callback time. |
| Extract Variables | Capture multiple fields from the call. |
| Voice Rescheduler | Handle rescheduling details after the conversation when configured. |
| Structured Output | Store call outcomes in a predictable format for review and reports. |
| Credentials | Allow approved tools to call protected systems when needed. |
| Analysis | Review what happened in the conversation and whether the assistant behaved as expected. |
Configure a Summary
- Go to Automation → AI Assistants → Post-Conversation Tools.
- Create or edit a summary tool.
- Define what the summary must include.
- Keep the summary short enough for an agent to read before a callback.
- Attach it to the assistant.
- Test a call and review the summary.
A good summary includes customer intent, important answers, objections, next step, and any handoff reason. It should not include unsupported assumptions.
Configure Variable Extraction
Use extraction when reports or follow-up workflows need structured values.
| Field | Example extracted value |
|---|---|
| Interest level | High, medium, low, not interested. |
| Budget | 75 lakh to 90 lakh. |
| Preferred location | Whitefield. |
| Visit intent | Wants visit this weekend. |
| Callback time | Tomorrow after 4 PM. |
| Outcome | Qualified, callback requested, wrong number, no answer. |
Keep field names stable. If managers compare reports over time, changing outcome labels every week makes analysis harder.
Configure Credentials Carefully
Credentials let tools connect to approved systems. They should be managed by admins or trusted operations users.
Before using credentials:
- Confirm the tool needs the credential.
- Confirm the credential belongs to the right account or integration.
- Confirm the tool only sends required information.
- Test with safe records before live use.
- Remove unused credentials from old tools or assistants.
Review Assistant Analysis
Use analysis pages after test calls and live calls to review:
- Whether the assistant followed the prompt.
- Whether it used the correct knowledge.
- Whether it captured required outcomes.
- Whether tool actions succeeded or failed.
- Whether the call should have been transferred earlier.
- Whether the final summary helps a human continue the work.
Troubleshooting Analysis Issues
| Problem | What to check |
|---|---|
| Summary is too generic | Add required summary sections and examples. |
| Extracted fields are missing | Ask for those details explicitly in the prompt or flow. |
| Outcome labels are inconsistent | Standardize allowed outcome values. |
| Tool action succeeded but report is unclear | Add structured output fields for tool result and final status. |
| Credentials fail | Review credential selection, permissions, and tool parameters with an admin. |
Manager Review Checklist
- At least one successful call summary is readable.
- Failed calls have useful failure reasons.
- High-intent calls are easy to identify.
- Follow-up owners know what to do next.
- Extracted values match what the customer said.
- Reports reflect the outcomes the business cares about.
Credentials are the admin-owned piece of this page — if a post-conversation tool needs to reach a protected system and the credential is missing, wrong, or failing, review it with an admin before changing anything else. The same goes for a hidden Post-Conversation Tools tab, which means your role doesn't cover these tools.
Contact Brixi support if an attached summary or extraction tool simply never runs — calls complete but produce no summary, fields, or outcome at all. Vague summaries and missing individual fields are configuration issues; fix those with the troubleshooting table first. When you do escalate, include the agent, the tool name, and a completed call that produced nothing.