Call and conversation workflow
Voice/text agent proof
A visible voice-agent pipeline with outbound test calls, transcripts, run state, and a dashboard for review.
Repo-local artifacts
Dashboard artifact
Repo-local screenshot showing the review surface for voice-agent runs.
Walkthrough video
Repo-local video artifact for the voice-agent MCP and dashboard workflow.
Workflow
How the workflow runs.
- 1
A test call can be started from the site modal through /api/voice-agent-test.
- 2
The route forwards the phone number and prospect name to the voice-agent testing endpoint.
- 3
The voice stack handles outbound Twilio calling, live audio, streaming transcription, and run state.
- 4
Runs, logs, transcripts, and summaries are reviewed in the dashboard.
What the owner sees
A dashboard record for the run.
Live or reviewable transcript content.
Operational logs and summary state.
A clear place to inspect what happened before taking action.
Human review point
The voice agent can collect and structure the call, but a person reviews the run before relying on it for screening, scheduling, or business decisions.
Connection to AI Audit
AI Audit identifies whether voice belongs in the workflow, where consent and review are needed, and what information should be captured.
View AI AuditConnection to launch work
A Launch Sprint can adapt the pattern into a bounded phone or text workflow with escalation rules, transcript review, and owner handoff.
View AI AutomationGrounded artifacts
- Repo-local dashboard screenshot in public/portfolio/voiceagentautomation.
- Repo-local MCP walkthrough video in public/portfolio/voiceagentautomation.
- Site API route for starting test calls against the voice-agent endpoint.
- Portfolio project copy describing Twilio calling, real-time speech bridge, transcripts, summaries, and MCP tooling.
Follow-up assets needed
- Add an anonymized example summary from a consented test call.
- Add a short public walkthrough clip if the dashboard view changes.
Bring your version