OpenAI’s Dot agent is enterprise software that can also order your dinner

OpenAI launched Dots this week, its latest AI agent platform that sits somewhere between helpful assistant and corporate productivity tool. According to The Verge, Dots come with customizable blob-like avatars and names, but the experience feels distinctly work-focused despite the approachable design.

The split-screen interface shows your conversation in one pane while the agent executes tasks in another. You can watch Dotty McDotface (or whatever you name yours) work in real time. Right now you're limited to one Dot, though OpenAI plans to support multiple agents eventually. The setup mirrors Meta's Muse, which launched last week, but the vibe is different. Where Muse aims for casual and conversational, Dots lean into getting things done.

This positioning matters more than it seems. OpenAI is clearly targeting business users who want AI agents that slot into existing workflows, not consumers looking for a chat buddy. The ability to order dinner feels like a feature added to prove versatility, not the core use case. That's smart. Enterprise customers pay subscriptions and need reliability, not just novelty.

The agent platform race is heating up fast. We've seen Anthropic's Claude with computer control, Microsoft's Copilot expanding across Office, and now Meta and OpenAI going head to head with consumer-facing agents. Each company is betting on a different interaction model. OpenAI seems to think people want clear separation between instructing the agent and watching it work, rather than a single unified chat experience.

The real test will be how well Dots handle complex, multi-step tasks without constant supervision. Ordering a burrito is table stakes. Can it manage a research project, coordinate with your calendar, and draft a presentation based on loose instructions? That's where productivity agents need to go to justify their existence beyond being chatbots with extra steps.

What this means for you: If you're using ChatGPT for work, Dots could become your persistent project manager rather than a one-off question answerer. Think of it as delegating a stream of related tasks instead of issuing individual commands. Try this prompt to test the concept with your current AI tool: "Act as my research assistant for the next hour. I'm preparing a market analysis on [topic]. First, outline what information we'll need. Then work through each section, and check in with me before moving to the next part." This trains you to think in agent mode, delegating workflows rather than individual tasks.