April 22, 2026
By Bob O'Donnell
The rise of AI-powered agents inside business organizations has quickly evolved from an intriguing possibility into an inevitable next step in enterprise computing. Not surprisingly, the companies building the technologies to power those agents—including Google—have accelerated their efforts accordingly. At last year’s Cloud Next event, Google focused on early adopters with the release of its Agent Development Kit and the debut of its A2A (Agent-to-Agent) protocol. Then, last fall, the company unveiled Gemini Enterprise, providing a framework and set of tools designed to help make agents a practical reality.
Now, at Cloud Next 2026, Google has taken a much bigger step forward. In fact, what the company revealed makes it clear that it is no longer simply offering a handful of agent-related tools. It is trying to build a broad enterprise platform for creating, deploying, managing, securing, and scaling agents across an organization. That is a big and ambitious move. At the same time, it also highlights one of Google’s biggest remaining challenges: making all these new capabilities digestible and accessible to a broader base of enterprise customers.
At the center of the announcements was Gemini Enterprise Agent Platform, an overhaul of the company’s Vertex AI toolset that brings together a comprehensive suite of capabilities for building, running, and securing agents in enterprise environments. Google also expanded Gemini Enterprise itself, positioning it as an entry point for both developers and end users who want to create and use agentic applications.
From an agent creation perspective, the company is clearly trying to broaden the audience. Developers can take advantage of the enhanced Agent Development Kit, while more advanced business users, comfortable with low-code and no-code tools, can leverage a new Agent Designer. Together, these offerings are intended to make it easier to build AI-powered agents that can automate workflows, perform specific tasks, and support a wide range of business processes. Google also created a dedicated section for agents in its Google Cloud Marketplace, where organizations can find agents built by Google as well as a growing range of third-party offerings from companies such as Salesforce, ServiceNow, and Oracle. Once created or acquired, these agents can then be deployed through the Gemini Enterprise application.
Of course, simply making it easier to build agents is not enough. Before most organizations are willing to deploy them broadly, several major concerns need to be addressed. Most notably, IT leaders remain worried about what happens when agent usage starts to scale across an enterprise. It is one thing for a small number of technically savvy users to create and use agents for their own work. It is something entirely different when thousands—or even millions—of agents begin operating across departments, applications, and workflows. At that point, issues such as governance, visibility, monitoring, and relevance become impossible to ignore.
To its credit, Google addressed several of these concerns directly. On the monitoring and management side, the company introduced Agent Identity, Agent Gateway, and Agent Monitoring capabilities to help organizations track what agents are doing, what information they are accessing, and how they are interacting with systems and data. Google also announced a new Agent Simulation tool that enables developers and IT teams to test various scenarios before agents are put into production. It remains to be seen how effective all of these capabilities will be in real-world deployments, but the broader point is undeniable: without these kinds of governance functions, widespread enterprise deployment of agents is unlikely to happen.
Google also tackled another critical challenge: context. As agents begin to handle longer, more complex workflows, which then get shared across teams, they need access to the right information in order to interpret requests accurately and act appropriately. At a basic level, the company is adding Memory Bank and Memory Profiles as part of an updated Agent Runtime engine to provide longer-term memory for agentic workflows. That should help improve continuity and make agents more useful over time.
At a deeper level, however, Google is also trying to solve one of the biggest enterprise AI problems of all: how to provide agents with consistent, organization-wide access to relevant data. In theory, that sounds straightforward. In practice, it can be extremely difficult, particularly when enterprise data is spread across multiple environments, stored in different formats and often locked into different clouds. Google’s new Agentic Data Cloud is designed to address those issues by allowing different data types, formats, and locations to be integrated into a unified knowledge framework. Built on an AI-native cross-cloud lakehouse architecture, Agentic Data Cloud enables organizations to keep data sources in place and avoid potentially expensive data egress charges that can come from moving data between clouds. That is a meaningful step, because if agents are going to be useful at scale, they need access to the right enterprise context without forcing organizations into costly and disruptive data migrations.
Security was another major theme, and for good reason. Many organizations are already dealing with agent-related security concerns as well as the spread of Shadow AI through browser-based tools and extensions. To help address those issues, Google announced a telemetry tool for monitoring AI workloads created by extensions in Chrome Enterprise, along with a Shadow AI reporting tool for Chrome that can provide greater visibility into browser-based AI activity. These can be used on their own or integrated into broader SecOps environments. Here again, the message is clear: if agents are going to become a core part of enterprise workflows, security and oversight have to be built into the foundation.
In addition to its enterprise-focused announcements, Google also introduced several agentic capabilities aimed at improving individual user productivity. Within Workspace, for example, the company has made applications such as Docs, Sheets, and Slides MCP-compatible, enabling them to participate in more sophisticated, agent-driven workflows. That is an important move, because it underscores a much broader shift in how productivity is starting to be defined. Instead of focusing on individual applications and isolated tasks, agents create the possibility of linking actions together around desired outcomes. Workspace Intelligence serves as the contextual glue that allows these types of connections to be seen and acted upon. Similarly, the ability to use Canvas mode in Google Enterprise (which then taps into the individual Workspace apps) provides a completely different way of thinking about how workflows can be created. There is still much more to come, but all these efforts represent a significant step toward a very different model of work. (See “The Outcome Economy: Surviving the Agentic Blitz” for more.)
Google has also integrated Gemini more deeply into Workspace applications. The inclusion of Google Gemini in Chat, for example, opens the door to scenarios where the system can analyze chat logs, determine that a presentation is needed, pull in current data from a Sheets spreadsheet, and then build a slide deck in a company’s preferred format and style within Slides. Because of the relatively open nature of the implementation, Google can also extend some of these capabilities across documents in Microsoft Office formats. For organizations considering a move from Microsoft’s productivity tools to Workspace, that could prove to be a surprisingly important advantage.
The company is extending these kinds of workflows beyond Workspace and into Chrome as well. New at Cloud Next was the ability to run agentic “skills” inside the browser, enabling more automated and more sophisticated multistep browsing workflows. That may sound like a small addition compared to some of the broader platform announcements, but it reinforces the company’s wider effort to bring agentic experiences into the everyday tools workers already use.
As is often the case at Google events, there was also a flood of news in other areas. From the debut of 8th generation TPUs—including new versions optimized for both training and inferencing—to the deeper integration of the company’s newly acquired Wiz security capabilities, Cloud Next once again demonstrated the unusually broad range of technologies Google can bring to market. That breadth matters. Google remains one of the very few companies that can design chips, build infrastructure, run a cloud platform, develop frontier models, offer data tools, provide security services, and integrate all of it into user-facing applications. In that sense, the company’s biggest competitive advantage may not be any one announcement, but rather the increasingly full-stack nature of its enterprise AI strategy.
Still, that same breadth can also create challenges. Given Google’s engineering heritage, the large number of new tools and services is not surprising. But as the company continues evolving into a provider of enterprise technologies for a much broader range of organizations—many of which are far less technically sophisticated than its early adopters—it needs to do more to make these offerings easier to understand, easier to adopt, and easier to use.
There is little doubt that advanced users and forward-looking organizations will find ways to drive meaningful productivity gains with the new tools Google and its competitors are bringing to market. But if the agentic revolution is going to have a truly broad enterprise impact, mainstream business users will need more than powerful technology. They will need tools that are approachable, understandable, and manageable in the context of everyday work. Cloud Next 2026 made it clear that Google understands where enterprise AI is headed. The next challenge is making sure the rest of the market can get there too.
Here’s a link to the original column: https://www.linkedin.com/pulse/google-pushes-agentic-ai-toward-enterprise-mainstream-bob-o-donnell-zqgrc
Bob O’Donnell is the president and chief analyst of TECHnalysis Research, LLC a market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on LinkedIn at Bob O’Donnell or on Twitter @bobodtech.
| 
