Agents Move Into Workflows

Google and OpenAI are moving artificial intelligence from chat windows into the software people already use to do work. (reuters.com) At Google Cloud Next in Las Vegas on April 22, Alphabet said AI agents are central to its enterprise push and folded key products under “Gemini Enterprise.” Chief Executive Sundar Pichai also reaffirmed Alphabet’s 2026 capital-spending plan of $175 billion to $185 billion, with “just over half” of machine-learning compute investment going to Google Cloud, Reuters reported. (reuters.com) Google Cloud Chief Executive Thomas Kurian said the “experimental phase” is over as customers shift from model demos to production systems. Google’s April 22 keynote introduced the Gemini Enterprise Agent Platform, including Agent Designer, an Inbox for agent activity, long-running agents, Skills and Projects. (reuters.com) (cloud.google.com) OpenAI made the same turn on April 22 with ChatGPT Workspace Agents for Business and Enterprise workspaces. The company said teams can build agents for repeatable tasks, test them before publishing, connect them to apps and tools, share them across a workspace, use them in Slack and run them on a schedule. (help.openai.com) That changes what vendors are selling. The pitch is no longer only a smarter model; it is software that can reach Slack threads, Salesforce records, files, approvals and schedules without employees copying information from one screen to another. (help.openai.com) (googlecloudpresscorner.com) The new battleground is control over those connections. OpenAI said workspace agents are off by default at launch for ChatGPT Enterprise, admins must enable them for eligible workspaces, and the feature is not available at launch for Enterprise customers with encryption key management, or EKM. (help.openai.com 1) (help.openai.com 2) OpenAI also tied the product to existing workplace channels instead of a separate assistant window. Its help center says Enterprise admins must enable the Workspace Agents in Slack app and can use role-based access controls to limit which users or groups get access. (help.openai.com) Google is making the same argument from the cloud side. Its April 22 partnership announcement with Salesforce said customers will be able to deploy agents in Slack and Google Workspace, with Agentforce and Gemini Enterprise supplying context and execution across both companies’ systems. (googlecloudpresscorner.com) Both companies are also framing scale as proof that this is now an infrastructure business. Google said nearly 75% of Google Cloud customers use its artificial-intelligence products, and more than 330 customers processed over one trillion tokens each in the past 12 months. (cloud.google.com) The next test is whether companies trust agents with real permissions, not just questions. In this phase of the market, the winners may be the vendors that can connect the most systems and still satisfy security teams, procurement teams and administrators. (reuters.com) (help.openai.com)