An AWS Example: From AI Experimentation to Mastering Agentic Commerce

All kinds of brands have spent the past few years figuring out what AI can do. It’s led to plenty of experimentation — some useful, some interesting, and some that would’ve been better kept on the meeting room floor.

In many ways, that dynamic was inevitable — new technology often tends to arrive before organizations have fully worked out where it best fits within their own operations.

But as AI becomes a more permanent part of the business, the question is starting to change: less so about what the technology can do in isolation, and more about what kinds of conditions businesses need to implement in order to fully unlock its potential.

For ecommerce brands, a lot of those conditions come down to the infrastructure behind the experience: the quality of the product information, the way that information moves through the organization, and whether teams are actually set up to use AI in a meaningful way.

And that’s a big deal when you consider the ways AI is already beginning to influence how shoppers find and evaluate products. As agentic commerce trends continue, these systems could take on an even larger role in deciding what gets surfaced, recommended, and eventually purchased.

In other words: smarter technology won’t compensate for a messy foundation. If product data is fragmented, inconsistent, or hard to trust, AI won’t solve the problem — it will make it worse.

From AI Experimentation to Business Transformation

For an unfortunate number of organizations, AI adoption has so far looked less like business or ecommerce transformation and more like accumulation. Teams start solving local problems with local tools, and suddenly the organization has a patchwork of assistants, automations, and agents without much of a bigger plan holding it together.

How AWS Found Synergy (Not Disparity) With AI Adoption

A “mod-podging” of AI applications was one of the problems faced by Amazon Web Services (AWS) as it expanded its own use of agentic AI.

According to Peter Cray, vice president of strategy and operations for AWS, more than 200 first-party tools and opportunity-to-cash processes were spread across more than 40 systems at the company. Adding more AI on top carried the very real risk of simply reproducing the same complexity the company was trying to escape in the first place.

Instead, AWS changed the question: Rather than asking where agents could be added, it chose go-to-market as a business domain to rethink more broadly. From there, the organization defined what it wanted the future state to look like, then worked backward to simplify the systems, data, and workflows required to support it.

It’s a useful lesson for ecommerce brands. Using generative AI to create thousands of product descriptions faster may save time, but if that content still has to move through disconnected spreadsheets, regional silos, and manual approvals, just how much organizational transformation is really taking place? It’s the same fractured model, just with a shiny new content engine attached.

Apply AI To Combat Bottlenecks and Reach Your North Star

The better starting point, according to Cray, is the business outcome itself. Which parts of the commerce operation are creating the most friction? Where are teams losing time? Where does inconsistent product information slow down launches or make it harder to support new markets and channels? Those are the places where AI can become more than an efficiency layer.

For global brands, success should be measured by what actually changes in the business — whether products get to market faster, teams spend less time wrestling with information, or the organization can grow without adding the same amount of operational friction. AI activity is easy to count; transformation is harder, but that’s the part that matters.

Build the Product Data Foundation for Agentic Commerce

Here’s a stat worth noting: According to Salsify consumer research, 22% of shoppers have purchased a product because an AI search tool or shopping assistant recommended it — up from just 5% in 2025. Worth noting: of shoppers who planned to spend more this holiday season, 30% are using AI chatbots like ChatGPT to discover gift ideas.

If AI is going to play a bigger role in the buying journey, product data becomes a much more visible part of the customer experience. Of course, winning the recommendation is one thing; surviving the follow-up is another.

AI and the Ever-Evolving Buyer’s Journey

A shopper might first encounter a product through an AI assistant, then check the brand site, a retailer listing, or another marketplace before buying. (It’s likely, as at least 27% of shoppers use AI for some purchases but verify with other sources, according to Salsify research.) If the details don’t match, or if important information is missing along the way, the customer’s trust can evaporate pretty quickly. 

Product data is no longer just the plumbing behind ecommerce; it increasingly determines:

• What AI systems can understand;

• What customers actually see; and

• Whether either one has enough confidence to keep moving toward a purchase.

That puts a lot more pressure on the information underneath the experience. If different teams, markets, or channels are working from different versions of the truth, AI doesn’t make that inconsistency magically disappear.

Build AI Into the Way Commerce Actually Works

Better product data gives AI something reliable to work with. But that only matters if brands also rethink what happens around it.

The real promise of AI isn’t just generating content faster or automating isolated tasks. It’s changing how work moves through the organization. AWS approached that problem by looking at where its employees were spending their time.

How AWS Implemented AI To Improve Efficiency and Relationships

Across roughly 35 roles and nearly 40,000 field employees, AWS broke work into its core functions to identify where agents could meaningfully reduce manual effort. Meeting preparation and administrative work emerged as areas where AI could take on more responsibility.

After those changes were implemented, one employee said AI had freed up more time for building relationships and solving customer problems instead of doing administrative work. The point wasn’t simply to make the same work happen faster, but instead, to change where employees were spending their attention.

That’s a useful way for commerce teams to think about AI, too. The goal isn’t to automate a task simply because it can be automated. It’s to look at how product information moves through the organization and ask where automation can remove friction without creating new handoffs, review headaches, or confusion about who owns what.

AI Must Fit in, Not Stand Out

McKinsey & Company notes that organizations leading on AI are nearly three times more likely to have fundamentally redesigned workflows rather than simply layering new technology onto existing processes. For ecommerce brands, that’s the difference between having AI capabilities and actually getting value from them.

The technology has to fit the way people work — and ideally make that work simpler. Otherwise, the organization hasn’t redesigned the workflow so much as added another thing employees have to deal with. That can make adoption a people problem as much as a technology problem. If employees don’t understand where AI fits into their work — or don’t see how it makes that work better — even a well-designed system can struggle to gain traction.

Prepare the Commerce Organization, Not Just the Tech Stack

Are your shoppers already changing faster than your commerce strategy? While agentic commerce may still be taking shape, the direction of travel is getting harder to ignore.

AI is already playing a larger role in how shoppers discover and evaluate products, and that puts more pressure on the systems sitting behind those experiences.

For brands, the opportunity is bigger than adding more tools. It’s about making sure product data is trustworthy, workflows are designed around how people and AI actually work together, and AI investments are tied to outcomes the business can measure.