Google’s agent push

Google used its Cloud Next conference on April 22 to recast AI agents as enterprise software that can take actions inside business workflows, not just answer prompts. (reuters.com) At the Las Vegas event, Google rolled out the Gemini Enterprise Agent Platform, which it described as the next version of Vertex AI for building, scaling, governing and optimizing agents. Reuters reported the company also folded products under a broader “Gemini Enterprise” name as it pushed enterprise customers as the clearest source of AI revenue. (cloud.google.com) (reuters.com) Google paired that platform with Workspace Studio, a no-code tool inside Google Workspace for turning routine office procedures into automated “skills” and flows. In separate Workspace posts on April 22, Google said teams can build invoice-review automations, add private Gems into flows, and use those skills anywhere Gemini appears in Workspace. (workspace.google.com) (workspaceupdates.googleblog.com) An AI agent is software that can plan a series of steps, call tools, and carry out work with less human supervision than a chatbot. Reuters said Thomas Kurian told the company’s conference that the “experimental phase” had ended as customers shifted from older machine-learning projects toward custom agents that can act on their own. (reuters.com) Google’s pitch was that once agents start acting inside email, documents, databases and security systems, the hard part becomes control. The new platform adds agent integration, DevOps, orchestration and security features, while Google’s networking group separately highlighted observability for agent traffic across clouds. (cloud.google.com 1) (cloud.google.com 2) The company also kept pushing Agent2Agent, or A2A, an open protocol for letting one agent discover, talk to and delegate work to another agent across different tools and vendors. The Linux Foundation said on April 9 that more than 150 organizations support A2A and that Microsoft and Amazon Web Services had both integrated it into their cloud agent platforms. (developers.googleblog.com) (linuxfoundation.org) That interoperability push addresses a basic enterprise problem: companies do not want every agent trapped inside one software stack. The Linux Foundation said A2A added multi-tenancy, security updates and signed agent identity features in version 1.0, while Google said its platform is meant to bring internal and partner-built agents into one managed system. (linuxfoundation.org) (cloud.google.com) Google backed the launch with adoption numbers meant to show this is already a business line, not a lab demo. The company said nearly 75% of Google Cloud customers now use its AI products, 330 customers processed more than 1 trillion tokens each in the last 12 months, and direct customer API traffic has climbed to more than 16 billion tokens per minute from 10 billion last quarter. (blog.google) (cloud.google.com) The competitive backdrop is straightforward: Reuters said OpenAI and Anthropic have also shifted resources toward enterprise accounts, while Google is arguing that its advantage is owning more of the stack, from chips and cloud infrastructure to productivity software. Sundar Pichai said Alphabet still plans $175 billion to $185 billion in capital spending this year, with just over half of machine-learning computing investment dedicated to the cloud business. (reuters.com) Google’s message in Las Vegas was that selling AI to businesses now means selling a system for supervising digital workers. The company opened Next by calling the “agentic enterprise” real and in production, and the rest of the launch was built around that claim. (cloud.google.com)