OpenAI just made agents you can stop watching.
The Summary
- OpenAI launched Workspace Agents, letting Business and Enterprise plan users build agents that work across Slack, Salesforce, Google Drive, Microsoft apps, and other enterprise tools without babysitting.
- These agents complete tasks even after you log off, drafting emails, pulling data, and sending reports autonomously across channels.
- Timing isn't random: OpenAI hired Peter Steinberger, founder of OpenClaw (the viral "AI that actually does things"), right before this launch.
- Available now on ChatGPT Business ($20/user/month), Enterprise, Edu, and Teachers plans.
The Signal
This is OpenAI's answer to the agent babysitting problem. Until now, most AI agents were glorified chatbots. You asked, they responded, you checked their work, you asked again. Workspace Agents break that loop. You give them a task, they go off and complete it across multiple apps and data sources, and they keep working even if you're offline. No hand-holding.
The architecture matters here. These agents live in ChatGPT but operate in Slack, Salesforce, Notion, Google Drive, and other enterprise apps. You can talk to them in Slack. They can pull data from Google Drive, draft an email in Gmail, and post a summary back to the Slack channel where you started. All without switching contexts or micromanaging permissions.
"It's the end of 'babysitting' agents and the start of letting them go off and get shit done for your business."
OpenAI gives two examples: an agent that scrapes product feedback from the web and sends a Slack report, and a sales agent that drafts follow-up emails in Gmail. These aren't hypotheticals. They're shipping today. The Agents tab in ChatGPT's sidebar now functions as a control center where teams discover, deploy, and manage shared agents. Think app store, but for autonomous workers.
The OpenClaw connection is the tell. Peter Steinberger's agent went viral because it promised to "actually do things" instead of just talk about doing things. OpenAI hired him, and weeks later this ships. That's not a coincidence. That's product strategy. OpenAI saw what resonated, acquired the talent, and scaled the idea to enterprise customers who've been begging for agents that don't need constant supervision.
Key differences from custom GPTs:
- Custom GPTs required manual input every time. Workspace Agents run tasks to completion autonomously.
- Custom GPTs lived inside ChatGPT. Workspace Agents integrate directly into third-party apps where work actually happens.
- Custom GPTs were consumer-friendly toys. Workspace Agents are enterprise tools with permissions, business logic, and compliance controls baked in.
This is OpenAI making a bet that enterprises will pay for agents that reduce headcount, not just assist existing headcount. A sales agent that drafts follow-ups in Gmail is doing work a BDR used to do. A product feedback agent that scrapes the web and reports to Slack is doing work a junior PM used to do. The pricing model ($20/user/month for Business plans) suggests OpenAI expects companies to treat these agents like team members, not features.
The Implication
If you're running a team that burns hours on repetitive cross-app tasks, this is your test case for the agent economy. Build one. Give it a real job. See if it actually reduces workload or just creates a new kind of overhead. The companies that figure out how to deploy agents without turning their Slack into a bot farm will have an edge.
Watch how OpenAI prices this as adoption scales. $20/user/month works when agents are supplements. It breaks when agents outnumber humans. If OpenAI shifts to per-agent pricing or usage-based billing, that's the signal they're serious about replacing roles, not just augmenting them.