Archive · 7 Oct 2026
AI Trade Agents Gain Ground While Humans Keep Signing
AI agents can help businesses find suppliers and negotiate prices.
Alibaba says its agent can also prepare a draft invoice.
But the agent cannot make the final business commitment by itself.
A person must approve important terms and confirm payments.
That helps keep responsibility with the business making the deal.
Payment companies and governments are also working on ways to identify agents and set rules for them.
The article says these limits can make AI useful before it is perfect.
It also suggests that proof of an agent’s authority should work across different platforms.
Alibaba.com says its AI sourcing agent can find suppliers, negotiate within buyer-set targets and draft proforma invoices.
Users must approve pricing, commercial terms and purchase orders, while payments require explicit confirmation.
Alibaba expanded Accio with Accio Work in March 2026; the platform handles business tasks across more than 100 markets.
Mastercard, Visa and Ant International are working on a framework to verify AI agents and link them to validated operators.
The article argues that defined authority and human accountability can enable useful automation while limiting risks.
- Who
- Alibaba.com and its AI sourcing platform Accio; payment networks Mastercard, Visa and Ant International are also developing agent verification.
- What
- AI agents are taking on sourcing and negotiation tasks, while humans retain approval of binding terms and payments.
- When
- The article describes developments through 2026, including Accio Work’s expansion in March 2026 and a sourcing toolkit added in July.
- Where
- Alibaba.com’s platform and trade across more than 100 markets; the article also discusses cross-border commerce.
- Why
- The approach aims to gain efficiency from AI while keeping business authority and accountability with a known human or operator.
- Accio monthly users
- Alibaba says Accio has more than 10 million monthly users.
- Markets served
- Accio Work handles tasks across more than 100 markets.
- Commerce task test
- Kuo Zhang reported that Alibaba’s strongest tested frontier model completed 61.7% of 107 real commerce tasks.
- Human approvals
- Users must approve pricing, commercial terms and purchase orders; payments require explicit confirmation.
- Agent verification
- Mastercard, Visa and Ant International began work on a proposed “Know Your Agent” framework in September.
- US standards effort
- The US National Institute of Standards and Technology launched an AI Agent Standards Initiative in February.
- Automated contracting
- The article cites UNCITRAL’s 2024 Model Law on Automated Contracting, which recognizes contracts formed through automated systems.