OpenAI just flipped the script on enterprise AI. On April 22, 2026, the company rolled out workspace agents in ChatGPT, Codex-powered bots that teams can share to tackle multi-step jobs. These aren't solo acts like the old Custom GPTs. They're built for groups, running in the cloud, handling reports, code, messages—even when humans log off.
Picture this: a sales agent drafts follow-up emails in Gmail. Or one that scours the web for product feedback and pings Slack with a summary. OpenAI's announcement spells it out. “Workspace agents are an evolution of GPTs. Powered by Codex, they can take on many of the tasks people already do at work—from preparing reports, to writing code, to responding to messages. They run in the cloud, so they can keep working even when you’re not. They’re also designed to be shared within an organization, so teams can build an agent once, use it together in ChatGPT or Slack, and improve it over time.”
Custom GPTs? Faded fast after their 2025 debut. Few stuck. OpenAI saw the gap. Workspace agents fill it, tying into organizational controls and permissions. Available now as a research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. Rollout gradual, per OpenAI's Business release notes. No extra cost mentioned yet. But Codex itself ties to credits in flexible pricing—workspaces buy them for heavy use.
And Codex. OpenAI's been pumping it up. Just weeks ago, on April 16, a massive update hit: background computer use, image generation, memory, browser integration. Agents now click, type, run plugins on Macs—without hogging your screen. Engadget reports it as groundwork for a super app merging ChatGPT, Codex, and browser. Thibault Sottiaux, Codex head, called it developer-focused for now. Broader rollout later.
From Solo Coder to Shared Operator
Codex started as code gen. Now? Full workflow beast. April 2 brought Codex seats—credit-based, no fixed monthly hit. Ideal for bursty tasks. Business notes confirm promo credits: up to $100 per new seat after first message, capped at $500 per workspace. Enterprise gets token-based rates; legacy plans stick to messages for now.
Teams build once. Share everywhere. Agents gather context, execute steps, seek approvals. Integrate Slack for pings, Gmail for drafts. The Verge highlights demos: web-scraping feedback bots, autonomous reporters. “OpenAI’s new workspace agents can be shared within organizations, so teams can build an agent once, use it together in ChatGPT or Slack, and improve it over time.”
But context matters. OpenAI's enterprise push ramps up. Same day, partnership with Infosys baked Codex into Topaz AI platform. TechCrunch notes it's part of Codex Labs—engineers embedding tools at clients like Accenture, PwC. Sam Altman tweeted Codex hit 4 million weekly users. Distribution network forming.
Reactions? Swift. On X, Henry Scott-Green, OpenAI agent product lead, posted: “Workspace agents are here!! I’m so excited to ship this.” Kimmonismus called it Codex as all-in-one app. firstadopter proclaimed 2026 the year of AI agents.
Challenges ahead. Agents need tight permissions—org controls help, but breaches loom. Costs stack with credits. Custom GPTs to agents conversion? Coming soon, OpenAI says. And competition heats. Google Cloud tossed $750 million at agent partners. Anthropic lurks.
Still. This shifts AI from chat to doer. Teams assign. Agents grind. Output flows. OpenAI bets big on Codex as the harness. If it sticks, work changes. Fast.
Early adopters test now. Business plans get first dibs. Watch rollouts. Metrics will tell if agents scale beyond hype.
WebProNews is an iEntry Publication