Uber Freight CTO: Focus on business fundamentals to win with AI in freight
Summary
At the Supply Chain AI Symposium 2026 in Chicago, Uber Freight CTO Val Marchevsky warned that many AI startups built as thin wrappers around fleeting capabilities have vanished, while companies that address core business fundamentals, such as legal AI firm Harvey.ai, endure. He argued that AI adoption in freight faces a higher bar due to costly failures, fragmented legacy systems, and a preference for human interaction, stressing that models work best in predictable, bounded environments where humans remain in the loop.
Marchevsky highlighted Uber Freight’s success in automating document processing, which eliminated over half of manual work through an ensemble of models rather than a single solution, and described the company’s shadow‑mode testing approach for new AI tools. He cited a widely reported 95% failure rate of AI proofs of concept, urging firms to involve process owners from day one, scrutinize data governance, and rely on internal benchmarks. Looking ahead, he sees promise in agentic software that integrates with legacy systems and in AI that can self‑correct its mistakes.
(Source:Freightwaves)