About the Client
FleetWorks is a Y Combinator–backed AI platform for the $1T freight brokerage industry. Its AI dispatcher, Sofia, matches independent trucking carriers with loads and handles the calls, texts, and emails that used to eat up their day.
About the Project
I owned FleetWorks’ carrier onboarding end to end, from research and conversation design to the data model, prototypes, and A/B test, turning a signup form into a chat with Sofia.
My Role:
Product Designer
Year:
2026
Skills Used:
Research, Conversation Design, Prototyping, A/B Testing
Process
Onboarding had one job: give Sofia , Fleetworks' AI agent, enough context to be useful from a carrier’s first session.
The Problem
The old signup asked for fleet size, then handed carriers off to a phone call with Sofia. Anyone who couldn’t talk right then landed on a load board that didn’t know them yet.
Iterating on the Wall
I printed every iteration and put it on the wall so the team could mark it up together. Early versions were still forms, asking carriers to build their own lanes, From → To.
The Four-Minute Phone Call
Then a Texas owner-operator spent four minutes on the phone explaining his lanes to me. Nobody was going to type that into a form. So we stopped asking for routes on day one and let Sofia learn them over time.
A Conversation, Not a Form
Signup became a short chat with Sofia: equipment, trip distance, where they like to run. Every answer shows up as “memory updated ✓,” skipping is fine (Sofia asks later), and it ends with a summary carriers can edit before the first call.
The Result
In an A/B test against the old flow, carriers who got the chat were 30% more likely to start using load search within 24 hours.
more likely to start using load search within 24 hours
What I’d Change Next
Fold the truck-count question into the chat so signup doesn’t open on a form, and split Sofia’s longer messages so they read faster on a phone.
Key Contributions
Owned carrier onboarding end to end, from research to A/B test
Ran carrier discovery calls and turned them into product decisions
Partnered with engineering on the data model behind the chat