Enterprise AI agents

Enterprise AI is moving from chat windows to software that keeps working after the prompt ends. On April 22, Google and OpenAI both rolled out new agent products aimed at teams, not just individual users. (techcrunch.com) (openai.com) OpenAI said its new workspace agents in ChatGPT are “Codex-powered” shared agents that can handle long-running workflows in the cloud and work inside ChatGPT or Slack. The company put the feature into research preview on April 22 for ChatGPT Business, Enterprise, Edu, and Teachers plans. (openai.com) Google used its Cloud Next ’26 event in Las Vegas on April 22 to introduce Gemini Enterprise Agent Platform, which it described as a system to build, scale, govern, and optimize agents. Google Cloud CEO Thomas Kurian said the Gemini Enterprise app is also getting Agent Designer, an Inbox for agent activity, long-running agents, Skills, and Projects. (cloud.google.com 1) (cloud.google.com 2) An agent is software that can pull information from connected tools, follow a sequence of steps, and take actions with approvals when needed. OpenAI’s examples include qualifying leads, drafting follow-up emails, updating customer relationship management records, and routing software requests into information technology ticketing flows. (openai.com) Google is splitting its pitch between technical teams and business users. TechCrunch reported Gemini Enterprise Agent Platform is geared toward information technology and other technical teams, while Google’s Gemini Enterprise app is aimed at employees who want agents for meeting scheduling, trigger-based processes, shortcuts, and file creation without switching apps. (techcrunch.com) (cloud.google.com) The common thread is that vendors are treating office work more like a workflow system than a chatbot session. Google’s own Cloud blog said customers are shifting from “a passive assistant” to “an active part of the team,” with specialized agents coordinating work across compliance, finance, supply chain, and older business systems. (cloud.google.com) That changes the enterprise buying list. OpenAI says admins can set permissions, approval checkpoints, monitoring, role-based access controls, and audit logs, while Google says its platform adds orchestration, DevOps, integration, and security on top of Vertex AI’s model tools. (openai.com) (cloud.google.com) It also creates work for partners that connect these agents to the software companies already run. On April 22, Google Cloud announced a $750 million fund for consulting firms, software vendors, and channel partners to speed up customer adoption of what it called “agentic AI.” (googlecloudpresscorner.com) Google is also leaning on model choice as a selling point. TechCrunch reported the new platform supports Google’s Gemini models and Anthropic’s Claude family, including Opus, Sonnet, and Haiku variants, as cloud vendors try to make agent platforms look less tied to a single model provider. (techcrunch.com) The near-term test is whether companies trust these systems with routine handoffs that employees usually do by copy-and-paste. Both launches were framed around the same bet: if the agent can stay inside company permissions and logs, the next step after chat is letting it do the work. (openai.com) (cloud.google.com)