Oct 1, 2026 · 47 min
Alibaba.com turns sourcing into agent-to-agent commerce
#244: How Alibaba.com Is Building the Future of Agent-to-Agent Commerce
The episode shows how AI agents could reshape global B2B trade while forcing companies to rethink product design, evaluation, and planning.
- 1Alibaba.com evolved from AI search into agents that manage sourcing, supplier communication, comparison, and fulfillment.
- 2Useful agents depend on the model, harness, and business context—not the underlying model alone.
- 3Commerce Agent Bench tests whether agents can complete realistic workflows, exposing gaps hidden by abstract benchmarks.
Don't miss
Kuo Zhang explains why Commerce Agent Bench evaluates agents in realistic business environments rather than relying on abstract model benchmarks.
The brief
Kuo Zhang, Alibaba.com’s president, describes a shift from AI-powered search toward agents that can handle the messy, multilingual work of global sourcing.
The product’s ambition is practical: help entrepreneurs define requirements, communicate with suppliers, compare responses, and manage fulfillment across a fragmented trade process.
Zhang’s framework—Model × Harness × Context—argues that capable agents require more than a frontier model; orchestration, guardrails, and business data shape useful behavior.
The larger bet is agent-to-agent commerce, where buyer, supplier, fulfillment, and demand networks connect through software acting on businesses’ behalf.
Alibaba.com’s Commerce Agent Bench tests agents inside realistic business workflows, highlighting why success on abstract benchmarks does not guarantee useful commercial work.