Field report SW-20 / M9 · sheet /operations
Operations & RestructuringFiled from the territory
AWS Marketplace lets AI agents buy software within set limits
Amazon’s AWS Marketplace agents can recommend software, check contracts and security evidence, and buy selected products within company-set budgets, while buyers retain approval controls.
- Grid
- SW-20 / M9
- Surveyed
- November 20, 2026
- Reading
- 4 min read
- Correspondent
- Olivia Hart

Field notes
- AWS Marketplace agents cover five procurement stages: discovery, contracts, vendor security, purchasing and setup.
- Companies can allow agents to make purchases 100% independently within approved categories.
- A June 2026 G2 survey found 39% of surveyed B2B software buyers named IT security reviews as the largest post-selection delay.
- Drata says one security review can cover 50 to 60 requirements distilled from a 100-line questionnaire.
- Vanta expects human approval for seven-, eight- and nine-figure deals.
By Olivia Hart
Amazon is rolling out AWS Marketplace AI agents that can purchase selected software and services without case-by-case human sign-off once companies authorize them. The rollout covers discovery, contract terms, vendor security checks, checkout and setup of supported software. Buyers can set budgets, designate permitted products and send other purchases to staff for approval.
How much purchasing can run without sign-off?
Matt Yanchyshyn, vice president of AWS Marketplace and partner services, told Fortune: “You shouldn’t have to approve that.” His example is a product manager who asks an agent to build a company website but may not know which outside sign-in vendor, such as Okta, to choose.
The agent can find a service that fits, then start a free trial and perhaps buy one month of pay-as-you-go use. That lets it build and test a working prototype without the approval hold that previously could stop deployment.
Will buyers still search for products by keyword?
“People aren’t searching with keywords anymore,” Yanchyshyn said, citing a shift AWS sees in its own metrics. Instead of typing “database,” buyers describe what they want to build and what the software must do; an agent turns that short prompt into a long list of requirements.
Marketplace can use that description to recommend a mix of software and services for the project. Once buyers have a product in mind, the checkout assistant explains its price, what the license permits and whether a purchase order may be required.
AWS is also adding a data category to Marketplace because AI agents increasingly use it. An agent evaluating a startup investment, for example, could pull funding data from Crunchbase.
Where does human approval still apply?
Yanchyshyn said companies set the autonomy boundary: “If they want a human handoff on everything, fine. But if they’re comfortable with a certain category of products to allow agents to make a purchase 100% independently, let’s let them.”
“Because the most important thing right now in this climate is speed,” he said. The buyer remains responsible for defining those guardrails.
An agent’s purchase authority does not automatically remove the need for due diligence. A company may let an agent buy a tool while its IT team checks how the vendor handles company data.
Can security reviews move faster?
A June 2026 survey by software review platform G2 found that 39% of surveyed B2B software buyers named IT security reviews as the largest source of delay after choosing a vendor. AWS is working with Drata and Vanta, two compliance automation providers, to let buyers review vendor security records directly within Marketplace.
Both providers help companies collect and share security and compliance evidence, placing that material inside the procurement workflow. For buyers, the practical gain is fewer handoffs between product discovery, compliance checks, purchasing and setup.
What changes inside a vendor review?
Adam Markowitz, Drata’s co-founder and CEO, told Fortune that one review can test a vendor against 50 to 60 requirements distilled from a 100-line questionnaire. “Now, with every vendor being an AI vendor, there’s a whole new set of criteria that security teams are assessing their vendors against,” he said.
Done manually, that review has required weeks of back-and-forth, Markowitz said. “Emails, requests, spreadsheets, questionnaires — the agent takes that down to minutes.”
An agent can retrieve documents a vendor shares, including a SOC 2 report, in which an outside auditor assesses the company’s security controls. It can check the report requirement by requirement and quote the exact supporting passage, allowing a reviewer to verify its conclusion.
Why speed matters now
Yanchyshyn said companies cannot afford months of clearance through security reviews, legal checks and budget approvals. AI tools can change from one week to the next. “People can’t wait three months to get access to the latest one,” he said.
Which purchases will stay human-led?
Bill Fitzgerald, Vanta’s vice president of global channel and alliances, told Fortune that executives want to expand AI use while governing the tools and agents inside their businesses. They ask: “Are we leaning in hard enough to leverage the productivity and the impact of AI to drive and scale our business?”
Fitzgerald expects agents to take on more small, high-volume purchases autonomously. For seven-, eight- and nine-figure deals, he said, “you’re always going to have a human in the loop.”
For procurement teams, the practical split is clear: agents can execute within defined categories, budgets and approval rules. Companies can reserve seven-figure and larger commitments for human approval and keep agent security conclusions open to reviewer checks. The immediate benefit is a shorter path from request to deployment, provided companies set controls before granting authority.
via fortune.com/feed/ (Source)
More from Olivia Hart
Show full bio
News editor covering regulation, policy and their impact on corporate strategy and chief-executive leadership at CEO Frontier.
474 articles