OpenAI launches Dots, a paid agent platform for business tasks

OpenAI has announced Dots, a new agent platform designed to function as a general-purpose digital worker for professional tasks. According to The Verge, the product is best understood as "Codex, but for regular people," positioning it as a workplace tool rather than a consumer assistant. The launch is gated to OpenAI's highest-tier subscribers, including the $100-per-month Pro account, distinguishing it from free competitors like Meta's Muse and Instinct.

The details

The interface mirrors other recent agent launches, featuring anthropomorphic avatars and a split-screen view where users chat with the agent in one window while observing its actions in a virtual machine in another. Currently, users are limited to a single Dot, though OpenAI states that multiple agents will be available in the future. The virtual environment comes pre-loaded with creative and development tools such as Blender and GIMP. A key differentiator is the ability to grant the Dot access to the user's local computer via the desktop ChatGPT app, allowing it to manipulate local files and applications directly.

In The Verge's testing, the agent demonstrated mixed results on consumer-facing tasks. When attempting to schedule an internet service provider installation, the Dot got stuck on a "human check" requiring a sustained mouse hold, forcing the user to intervene manually. It also failed to locate a free trial option for a coworking space, instead offering a $35 day pass, while a competitor agent found the free option immediately. Security checks posed a recurring hurdle; the Dot was blocked by looping security checks on an IKEA account and a local restaurant's ordering page, issues that The Verge notes did not occur with Muse or Instinct.

However, the agent performed significantly better in controlled, professional environments. When given access to a personal website, the Dot successfully iterated on a new design based on voice instructions provided over a 10-minute phone call. It also handled a complex local task: consolidating multiple video files on a desktop into a single, correctly sized social media clip. The Verge notes that this required a "shocking number of permissions" but resulted in a functional output that the user would have struggled to produce efficiently on their own.

Context

Dots enters a crowded market of AI agents that have emerged over the past year. Meta's Muse and the buzzy competitor Instinct are currently free, positioning themselves as personal shoppers and general assistants. OpenAI's decision to gate Dots behind a premium subscription signals a different strategic intent. As The Verge observes, the gated rollout "says a lot about who the product is for," suggesting a focus on enterprise and professional users who are willing to pay for reliability and privacy over free, ad-supported convenience. This aligns with OpenAI's broader push into developer tools, where Dots can be seen as a consumer-facing extension of its coding agents.

What it means for developers

For developers and technical professionals, Dots offers a new way to offload repetitive digital labor. The ability to use the agent for local file manipulation, such as video editing or website deployment, could streamline workflows that currently require manual intervention. The integration with local desktop environments is particularly notable, as it bridges the gap between cloud-based agents and local machine resources. However, developers should be cautious with the permission model. Granting an AI agent access to local files and browser sessions requires a high level of trust in OpenAI's security and the agent's decision-making. The Verge's experience suggests that while the agent is capable, it still requires significant human oversight, particularly when interacting with external websites that employ bot detection.

What to watch

• Multi-agent support: OpenAI has promised the ability to have multiple Dots in the future, which could allow for parallel task execution and specialized agent configurations.
• Security improvements: The frequent failures on "human checks" and security loops suggest that OpenAI needs to improve its browser fingerprinting and interaction logic to compete effectively with free alternatives.
• Enterprise adoption: The $100/month price point will be a barrier for many, but success in the Pro tier could drive broader enterprise adoption, potentially leading to new B2B features.
• Competitive response: Meta and other players may accelerate their own agent development or introduce paid tiers to counter OpenAI's premium positioning.