AWS CEO Matt Garman says enterprise AI is moving from experiments to ROI

AWS CEO Matt Garman is making a direct push for an “outcomes first” view of enterprise artificial intelligence, telling executives that the current wave of AI deployments is starting to show measurable returns rather than remaining trapped in pilots.

In a new interview published on Amazon’s newsroom site, Garman frames the market transition as one in which businesses are moving past experimentation and toward projects that connect AI to business results. The company’s public messaging emphasizes a shift from trial-and-error toward implementations that can be evaluated for real-world value.

The interview also positions the AWS platform as a key enabler of that transition, implicitly contrasting the early era of AI hype with a more practical phase in which enterprises are expected to manage costs, reliability, and integration needs as they scale AI across teams and workflows.

While the newsroom release does not provide detailed figures or named customer case studies in the material available for this report, it does describe the core theme: enterprise AI is reaching a point where leaders are focusing on returns on investment. That emphasis matters because it indicates that AWS is leaning into AI not only as a technology upgrade, but as a business capability that must compete with traditional software on measurable performance.

For Amazon, the message is timely. AWS has been positioning its cloud offerings to include a growing layer of AI capabilities, and the company has repeatedly highlighted demand from enterprises seeking production-ready ways to deploy machine learning and generative AI. Garman’s remarks fit that broader narrative by focusing on how organizations determine whether AI is delivering value.

Enterprise AI “ROI” is a catch-all term, but in practice it typically means that AI initiatives are expected to show benefits such as lower operating costs, faster service cycles, improved decision quality, or increased revenue and retention. In AWS’s framing, the goal appears to be helping businesses reach that measurement stage, where deployments can be justified and expanded.

Even so, the interview’s publishable details in the available material are limited, and Amazon does not spell out the specific drivers of ROI in a way that can be verified here. The newsroom page presents the interview as a qualitative explanation, without supplying metrics, timelines, or concrete examples of which AI use cases are producing the strongest returns.

Still, the direction of travel is clear: AWS is treating enterprise AI as an operating model change, not just a model innovation. As companies move from proof-of-concept work to production deployments, topics such as governance, integration with existing systems, and sustaining performance become central, and those are areas where a cloud platform can differentiate.

Going forward, investors and customers will likely look for follow-through in the form of quantified results, named deployments, or more explicit benchmarks tied to cost and performance outcomes. The next tell will be whether AWS can translate the “real returns” message into repeatable, externally verifiable evidence as more enterprises scale AI beyond early experiments.