Agents become the enterprise battleground
- OpenAI launched subscriber 'workspace agents' and Google unveiled an AI agent suite at Cloud Next, both targeting workflow automation. - Google pitched Workspace Studio, the A2A protocol and Project Mariner as a full‑stack agent play against OpenAI. - The product race is shifting from model benchmarks to who owns orchestration, integrations and persistent team workflows (siliconangle.com) (thenextweb.com).
OpenAI and Google both spent April 22 turning AI agents into workplace software, not just chatbots. (openai.com) (blog.google) OpenAI said its new workspace agents in ChatGPT can handle repeatable tasks, run in the cloud, and be shared across a company inside ChatGPT or Slack. The company said the agents are powered by Codex and work inside an organization’s existing permissions and controls. (openai.com) OpenAI’s help center said the rollout is starting over the next few weeks for ChatGPT Business and Enterprise workspaces. It said eligible customers can connect apps including Google Drive, Google Calendar, Slack, and SharePoint, then schedule agents to run automatically. (help.openai.com) Google used Cloud Next 2026 to pitch a broader stack for the same job. Its event recap highlighted a Gemini Enterprise Agent Platform and described “the agentic enterprise” as the company’s roadmap for organizations that want software to complete multi-step work. (blog.google) (cloud.google.com) That changes the contest from whose model scores highest on tests to whose software sits inside everyday work. OpenAI is tying agents to ChatGPT, Codex, Slack and connected business apps, while Google is tying them to Google Cloud, Google Workspace and its developer tools. (openai.com) (cloud.google.com) Google’s case is that an agent is only useful if it can talk to other agents and company systems. Google Cloud said its Agent2Agent, or A2A, protocol is an open standard for agents to collaborate across platforms, and said more than 150 organizations are supporting the effort. (developers.googleblog.com) (cloud.google.com) The company also used Cloud Next to push products around that protocol, including Workspace Studio and Project Mariner, as part of a full-stack pitch that stretches from chips and models to inboxes and browser actions. That framing was central to Google’s event coverage and outside reporting from Las Vegas on April 22. (thenextweb.com) (blog.google) OpenAI is making a parallel argument from the other direction: start with the tool workers already open all day. Its product post said workspace agents are “an evolution of GPTs,” built for shared team use and long-running workflows rather than one-off prompts. (openai.com 1) (openai.com 2) The practical fight is over orchestration, which is the software layer that decides what an agent should do, which app it should call, and when it should hand work to another system. Google’s protocol push and OpenAI’s connected-agent rollout both target that layer, because it is where recurring business processes actually live. (developers.googleblog.com) (openai.com) Both companies are now selling the same promise in different packaging: software that can keep working after the user closes the tab. The next test is whether customers adopt the agent layer that comes with ChatGPT or the one bundled into Google’s cloud and productivity stack. (openai.com) (blog.google)