Skip to content
Advertisement
Amazon lets AI agents buy software on AWS Marketplace

Amazon lets AI agents buy software on AWS Marketplace

Amazon logo (public domain) via Wikimedia Commons

New procurement agents can research, vet and purchase software within budgets and approval rules set by companies

Amazon wants AI agents to take over one of the least glamorous jobs in corporate life: buying software.

The company is rolling out a suite of AI agents on AWS Marketplace that help businesses purchase software and procure services, Fortune reported. The standout feature lets companies pre-authorize agents to carry out certain actions without waiting for a human to sign off.

What the agents actually do

The update spans the full procurement journey. Both people and agents get tools for finding products, reviewing contract terms, checking a vendor’s security, completing the purchase and setting up supported software.

Companies can assign an agent a budget for specific products and write rules deciding which purchases get kicked to a person for approval.

In practice, that could let a team allow an agent to handle routine buys on its own while sending larger or unusual purchases to a manager.

How AWS built up to this

The launch is the latest step in a steady push to make AWS Marketplace agent-friendly. AWS introduced an “AI Agents and Tools” category in the marketplace on July 16, 2025.

Advertisement

The original plan reportedly called for around 50 listings. The category instead opened with more than 800 partner solutions, a sign that vendors were eager to get in front of AWS customers.

The growth kept going. Mid-2026 tallies put the category at more than 2,100 listings in one count and more than 3,400 in another, well beyond the initial target by over 40x on the lower figure.

Customers can discover, buy and deploy pre-built agents and related tools through standard procurement channels. Deployment options include APIs and containerized setups through Amazon Bedrock AgentCore, AWS’s infrastructure for running agents.

In November 2025, AWS added “agent mode,” a conversational interface for shopping the marketplace. Buyers can describe what they need in plain language, upload documents, get AI-generated summaries and build purchase proposals they can customize.

Agent mode was made available across more than 30,000 listings. It leans on semantic search, which matches on meaning rather than exact keywords, and the Model Context Protocol (MCP), a standard for connecting AI models to outside tools and data.

The commercial payoff showed up quickly. The agent mode enhancements reportedly doubled marketplace conversion rates, from approximately 3% to 6%.

AWS kept layering features into 2026. Those included flexible pricing models, automated private offers that produce customized pricing, and governance tools such as Amazon WorkSpaces for AI agents.

AWS executives have pointed to a 60% reduction in procurement time and 30% faster deployment for marketplace solutions.

What this means for buyers, sellers and AWS

For enterprises, the control point moves. Instead of approving each purchase one at a time, finance and IT teams will increasingly approve the rules: the budgets, the product lists and the thresholds that trigger human review.

For software vendors, listings, pricing and contract terms will likely need to be legible to agents, not just persuasive to people, since an agent is the one doing the first round of comparison shopping.

For AWS, a marketplace that converts better and closes deals faster keeps more spending inside the AWS ecosystem, and agents with standing budgets are, almost by definition, repeat customers.

The pitch also targets buyers with heavy oversight requirements. The combination of pre-authorized spending and mandatory human review for certain purchases is designed for public-sector and enterprise clients that run complex procurement but still want a person in the loop.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.