{{item.videoDuration}}
{{item.title}}
{{item.text}}
{{item.videoDuration}}
{{item.text}}
From agentic platforms to infrastructure to governance, here’s what stood out—and what it means for retail’s AI moment.
Much of the discussion around agentic commerce to date has centered on the front-end shift, where AI agents browse products, draw comparisons, and even execute transactions on behalf of human customers. That shift has big implications for retailers, ushering in an era in which they must market to and optimize for agents as much as for human browsing.
But what about the internal plumbing that will enable retailers to step into this new reality? From our vantage point at PwC, many retailers are still underestimating the operational lift required, from data readiness to governance to change management. At Google Cloud Next 2026, Google Cloud made it clear that it wants to assist enterprises along that journey. The hyperscaler debuted a series of new products built to help organizations establish and manage the underlying infrastructure needed to do agentic AI well.
The larger takeaway: the agentic era has arrived, and it’s time for organizations to move beyond experimentation. Agentic commerce sits within a larger rise of the “agentic enterprise,” a term we heard often at Next. If agents are going to negotiate pricing, manage inventory decisions, and execute transactions on their own, businesses must engage in real infrastructure work. They need an operating model where every agent has an identity, every action leaves a trail, and every data call hits governed, contextualized sources.
Toward a platform approach and the “agentic enterprise”
Retailers are heading toward a world where dozens of AI agents, perhaps many more than that, underpin their operating model. Each of them will be tasked with discrete workflows, from pricing optimization to digital merchandising to customer service. But without a centralized way to manage them, the ecosystem quickly becomes unworkable. You need a way to kick off agents, connect them to other systems, monitor their results, and continuously optimize.
With the Gemini Enterprise Agent Platform, announced at Next, Google Cloud is building that scaffolding, an endeavor we’ve also taken on at PwC. Agent Platform is a new, comprehensive platform to build, scale, govern, and optimize agents. It’s the evolution of Vertex AI, bringing the model selection, model building, and agent building capabilities that customers love, together with new features for agent integration, DevOps, orchestration, and security.
Some early enterprise deployments of Google Cloud AI agents—shared by Google in recent weeks—have shown us what this can look like in practice:
The Home Depot—Magic Apron: The suite of generative AI tools helps customers find answers that guide home improvement projects by combining a proprietary knowledge base with large-scale datasets and The Home Depot product info. Built for DIYers and pros alike, it can do everything from walk a customer through a project step-by-step to identify the best products for the job—down to the exact store location and aisle where they’re available.
Macy’s—Ask Macy’s: Built in just four weeks using Gemini Enterprise, Ask Macy’s is a conversational AI agent that helps customers navigate product discovery, answer detailed questions, and surface personalized recommendations in real-time. It centralizes text, images, and virtual try-on in a single multimodal experience. Connected to Macy’s product catalog and its customer data, the agent acts as a digital sales associate to guide decisions, reduce friction, and drive conversion.
Virgin Voyages—Rovey: Rovey is an AI agent designed as a personal concierge, helping cruise travelers with insights that are proactive, contextual, and brand-voiced. It’s more of a “crew assistant” than a traditional chatbot. Rovey supports customers as well as internal teams. That means it can help with everything from itinerary planning and booking inquiries to operational workflows behind the scenes.
Solving the data problem
These agent-driven innovations only work if the AI has access to the right data. Google Cloud also announced the Agentic Data Cloud at Next. Rather than forcing organizations to rebuild their data architecture from scratch, Google Cloud is aiming to unify fragmented data sources and make them usable and AI-ready in real-time.
The solution goes a step beyond the traditional data lake by not only aggregating data from across an organization, but also enriching it with context. For retailers, above all else, that means customer context, the accumulated understanding of who a customer is, how they shop, what they’ve bought, what they’ve returned, and what they’re likely to want next.
In an agentic world, customer context is a moat. Like the models themselves, agent platforms will become increasingly commoditized. But competitors can’t replicate a retailer’s proprietary, contextualized view of its own customers—or the agents that operate on top of it.
Governing agents at scale
As retailers grant agents increased autonomy, the stakes around security and governance are rising. Most organizations have some level of governance in place, but they often fall short of the frameworks they’ll need in an agentic environment.
Organizations must manage identities and permissions, for instance, while layering in accountability across a growing web of agents. They also must establish clearly what agents are allowed to access and do. With their announcements at Next, Google Cloud is aiming to make these considerations easier for enterprises by establishing a shared governance layer.
And then there’s security, presented as a core capability of Google Cloud’s platform approach, rather than a bolt-on, and underscored by two new products. Agentic Defense is Google Cloud’s cybersecurity platform, developed with Wiz to protect AI apps and workflows. Fraud Defense, meanwhile, extends fraud protection to agent-initiated transactions. That’s harder than it sounds, because the signals fraud teams have relied on for two decades—like device fingerprints, behavioral biometrics, and session anomalies—were built for humans clicking buttons. When the real shopper is an agent acting on a customer’s behalf, retailers need new ways to verify intent, authorization, and authenticity.
Retailers must be able to define guardrails and ensure compliance across the entire commerce lifecycle, from product discovery to payment. And they must have a clear view into the costs and value of their deployed agents. This level of control and transparency will separate organizations that scale agentic AI successfully from those that stall out during experimentation.
What comes next for retailers
Right now, retailers are balancing competing impulses. They face competitive pressure to deploy and get business value out of AI quickly. But speeding ahead without establishing a foundation is a recipe to waste money and fall behind.
With their announcements at Next, Google Cloud is aiming to address the challenges in setting that foundation. Regardless of cloud provider, retailers that lead the charge toward agentic commerce won’t be the ones attempting to boil the ocean. Instead, they’ll focus on building the underlying capabilities: clean, accessible data, presented in context; strong, trustworthy governance; and the ability to deploy and iterate agents quickly.
To learn more, register for our upcoming webcast on the rise of the agentic enterprise platform, including a Q&A with Google Cloud on the future of agentic commerce.
Missed us at Google Cloud Next?
{{filterContent.facetedTitle}}
{{contentList.loadingText}}
{{contentList.loadingText}}
Follow us
