
The problem: brokerage runs on phone calls and email
A freight broker's day is a long list of calls and emails: finding a carrier for a load, agreeing a rate, checking where the truck is. Most of that work is repetitive, time sensitive and expensive to do by hand. LoadBetter wanted AI agents that take over the repetitive conversations while brokers keep control of the decisions.
Voice agents for carriers and drivers
Voice agents built on Vapi place and answer calls triggered by business rules: booking calls with carriers, tracking calls with drivers and outbound call campaigns. Every call is tied back to the load in the platform, so the broker sees what was said and what changed.
An email agent that negotiates rates
The email agent reads inbound carrier emails, extracts the offered rate, drafts a counter offer and can send it automatically. Real traffic is messy (rates written without a currency sign, or as "1.7k"), so every miss found in production became a regression test. Signed reply tokens are shared between the TypeScript API and the Python service so a reply always lands on the right negotiation.
Built to be trusted in production
- LLM guardrails with deterministic fallbacks when the model is unsure
- Human in the loop flows for the decisions that need a broker
- Covered by SpotBot's suite of about 4,200 automated tests, from unit tests to end to end runs and replayed voice scenarios
- A unified driver login (phone and SMS code) across the API and the Expo driver app, plus "Sign in with LoadBetter" as an OAuth2 provider