Enterprise AI shifts to agents
- Cloud vendors are shifting focus from raw models to agent platforms that act inside workflows, not just answer queries. - Google, OpenAI and others unveiled tools that integrate agents with Slack, Salesforce, Gmail and enterprise controls. - The market is now about which environment will manage agent governance, integrations and data access rather than just which model to use (reuters.com).
Big cloud vendors are now selling AI agents that do work inside business software, not just chatbots that answer questions. (reuters.com) At Google Cloud Next in Las Vegas on April 22, Alphabet said AI agents are central to its enterprise push and folded Vertex AI into a new Gemini Enterprise Agent Platform. Google also said nearly 75% of Google Cloud customers use its artificial intelligence products, and 330 customers processed more than 1 trillion tokens each over the last 12 months. (reuters.com) (blog.google) OpenAI made a similar move on April 22, introducing workspace agents in ChatGPT for Business, Enterprise, Edu and Teachers plans. OpenAI said the Codex-powered agents can run long workflows in the cloud, be shared across a company, and work in ChatGPT or Slack under organization-set permissions. (openai.com) An AI agent is software that can pull context, follow steps, ask for approval and take actions across tools. That is different from a standard chatbot, which mainly responds to one prompt at a time and waits for the next one. (openai.com) (reuters.com) The new fight is moving up a layer from the model itself to the system that decides what data an agent can see, what apps it can touch and how its actions are logged. Google said its new platform combines model choice with integration, orchestration, security and governance features for agents. (cloud.google.com) (reuters.com) Google paired that platform with Workspace Intelligence, a system that grounds Gemini in Gmail, Chat, Calendar, Drive, Docs, Sheets and Slides, with new admin controls over which data sources it can use. Google also announced new “skills” in Workspace Studio so teams can turn standard operating procedures into reusable automations. (workspaceupdates.googleblog.com) (workspace.google.com) Salesforce is making the same case from the control side. On April 15, Salesforce expanded Agent Fabric with centralized large language model governance, monitoring dashboards and discovery tools for agents and Model Context Protocol servers across platforms including Amazon Bedrock and Microsoft Foundry. (salesforce.com) Google and Salesforce then tied their systems together on April 22. The companies said customers will be able to deploy agents in Slack and Google Workspace, with Agentforce and Gemini Enterprise sharing context across both platforms for end-to-end workflows. (salesforce.com) Reuters reported that Google is trying to prove returns on heavy generative artificial intelligence spending, with Sundar Pichai reaffirming Alphabet’s plan to spend $175 billion to $185 billion in 2026 and saying just over half of machine-learning computing investment will go to the cloud business. Thomas Kurian told Reuters that Vertex AI’s main use has shifted from older machine learning work to a surge in custom agent building. (reuters.com) The sales pitch is no longer just “our model is smarter.” It is “our platform can let agents work across Gmail, Slack, Salesforce and internal data without losing security, approvals or audit trails.” (openai.com) (cloud.google.com) (salesforce.com)
