Amazon's updated Business Solutions Agreement, effective 4 March 2026, introduced an Agent Policy for automated software and AI agents that access Amazon services. As an operator that uses AI agents for advertising, promotions, inventory and listings, we read it closely.
What the policy asks
- Agents must identify themselves as automated systems.
- Agents must comply with the Agent Policy.
- Agents must stop accessing Amazon services when Amazon requests it.
The agreement also restricts using Amazon materials to develop AI and machine-learning systems. Sellers have asked where exactly the line falls for common tools, and Amazon has not published thresholds, so we take a conservative reading.
How we work within it
- Approved routes first. Our agents work from Amazon's official reports, the Selling Partner and Advertising APIs and approved third-party tools, not from scraping.
- Identified and logged. Every automated action is attributable to a named agent and recorded with its inputs, so we can show what was done and why.
- Human approval. Agents prepare recommendations; people approve changes to prices, listings and anything outside agreed limits.
- A clean stop. Each agent can be paused without affecting the rest of the operation, which is what a request to stop requires in practice.
Why this matters to partners
Brands that hand their channels to an operator are entitled to know that the tools used on their accounts are within the marketplace's rules. Automation is only an advantage if it does not put the account at risk.