AI voice freight-broker assistant
Zuum
A production voice agent integration to already developed system that calls truck drivers for check-ins and shipment status, extracts structured load data from every call and keeps dispatch up to date in real time.
- Interruptible voice calls
- Real-time
- Structured data from every call
- JSON
- What we built
- Voice agentDispatch dashboardPrompt workspace
The problem
Freight brokers spend a large part of every day on the phone with drivers: confirming pickups, checking ETAs and chasing proof of delivery. The calls are repetitive but time-sensitive, and the answers end up scattered across notes, inboxes and spreadsheets.
What we built
We built a voice assistant on Retell AI that places check-in calls and handles real, interruptible conversations with streaming speech recognition and synthesis. Slot-filling and guardrails reliably turn each call into structured JSON: ETA, current location, exceptions and POD requests.
A Python backend with REST endpoints, webhooks and WebSocket updates orchestrates calls, stores results in Supabase and triggers alerts, with retries and idempotency around every telephony event. A web dashboard shows live calls, transcripts and QA feedback, and a prompt workspace lets the team change the agent's behaviour, output schema and safety settings without a deploy. Call outcomes, completion rates and latency are tracked to guide improvements.
Key features
- 01Interruptible, natural conversations with streaming ASR and TTS
- 02Check-in workflows across the journey: origin, transit and destination
- 03Slot-filling and guardrails that extract ETA, location, exceptions and POD requests as JSON
- 04Live call monitoring with transcripts and QA feedback
- 05Prompt and configuration workspace to iterate on voice behaviour and schemas
- 06Webhooks, retries and idempotency around every telephony event
- 07Alerts and notifications when a load needs attention
- 08Analytics for call outcomes, completion rates and latency
Inside the product.
2 images
Tech stack
- Voice AI
- Retell AITwilio
- AI
- OpenAILangGraphLangChain
- Backend
- PythonFastAPIWebSocketsSQLAlchemy
- Data
- SupabasePostgreSQL
- Web
- Next.jsReactTypeScriptTailwind CSS
- Platform
- DockerAWS
Results
- Dispatchers spend their time on exceptions instead of routine status calls.
- Shipment data stays fresh and accurate without manual entry.
- A reusable platform for further voice-driven logistics workflows.
Next project
Vesta AI
Free, no obligation
Have something similar in mind?
Book a free consultation and we'll conduct a meeting to talk it through, or start with a free audit. No obligation either way.