Google bets on agents

Google used its Cloud Next conference in Las Vegas on April 22 to make AI agents the center of its enterprise sales pitch. (usnews.com) Alphabet’s cloud unit paired that message with a $750 million fund for partners building and deploying agent systems. Google said the money will go to consulting firms, systems integrators, software partners and channel partners across its 120,000-member ecosystem. (googlecloudpresscorner.com) Google also folded its enterprise artificial intelligence tools under the “Gemini Enterprise” name and introduced the Gemini Enterprise Agent Platform as the successor to Vertex AI’s earlier agent-building setup. Google Cloud said the platform adds orchestration, DevOps, integration, governance and security features for agents. (usnews.com) (cloud.google.com) An AI agent is software that does more than answer a prompt once: it can plan steps, call tools, connect to company systems and complete a task with less hand-holding. Google Cloud Chief Executive Thomas Kurian told Reuters that Vertex AI’s main use case has shifted from older machine-learning projects to customers building custom agents. (usnews.com) That pitch depends on another layer most buyers never see directly: embeddings, which turn text, images, audio or video into numeric fingerprints so software can find related material by meaning instead of exact keywords. Google made Gemini Embedding 2 generally available this week on Vertex AI after first releasing it in public preview on March 10. (docs.cloud.google.com) (blog.google) Google says Gemini Embedding 2 accepts text, images, audio, video and PDF documents, maps them into one shared vector space, and supports more than 100 languages. The model outputs vectors up to 3,072 dimensions and is aimed at retrieval, classification and clustering workloads. (ai.google.dev) (docs.cloud.google.com) For enterprise customers, that means the buying conversation is moving from “which chatbot should we use” to “which platform can search our data, route work across systems and keep the whole flow governed.” Google’s own product pages now describe Gemini Enterprise as an end-to-end system for “the agentic era,” with no-code tools for building assistants and more complex orchestrators. (cloud.google.com 1) (cloud.google.com 2) The competitive backdrop is straightforward: Reuters reported that OpenAI and Anthropic have also shifted resources toward business customers, while Google is trying to close ground on Amazon and Microsoft in cloud infrastructure. Google is arguing that its advantage is a full stack that runs from custom chips and data centers up through models, retrieval and workplace software. (usnews.com) (blog.google) The immediate test is whether companies pay for systems that can be trusted with real workflows, not just demos. Google’s announcements this week were built to show that agents, retrieval and partner-led deployment are now one enterprise product sale. (cloud.google.com) (googlecloudpresscorner.com)