OpenAI releases dots, its always-on personal assistant

OpenAI has officially launched dots, a new always-on personal assistant designed to operate in the background even when users are not actively prompting it. According to TechCrunch, the product was unveiled at OpenAI’s developer day in late September and is currently available to subscribers on the Pro tier or higher, with plans starting at $100 a month. The launch marks a significant shift in how OpenAI is positioning its technology, moving away from discrete task-based interactions toward a persistent agent that continuously monitors user context, calendars, and emails to proactively suggest actions.

Alexander Embiricos, who leads OpenAI’s enterprise product team, explained the strategic pivot to TechCrunch at Disrupt this month. He noted that the company realized that framing the AI as a single entity with one continuous conversation, rather than many separate chats, made the technology feel more intuitive. "There’s kind of this convergent evolution happening," Embiricos said, describing the industry-wide trend toward persistent assistants. The name "dots" is a deliberate play on the negative space inside OpenAI’s logo, intended to feel ubiquitous and less technical than previous offerings.

From Coding Tools to Personal Assistant

The core functionality of dots is not new to OpenAI’s stack. Embiricos acknowledged that much of what dots does can be achieved with OpenAI’s Codex app, its agentic coding tool released earlier this year. However, Codex requires significant manual oversight and "hand-holding" to function effectively. Dots aims to remove that friction by automating the monitoring and decision-making processes.

The assistant connects to a user’s calendar, email, computer, and other tools to identify potential issues before they become critical. In our testing, the assistant’s ability to think ahead is apparent and often useful, though the timing of its alerts remains a work in progress. Embiricos shared an anecdote about his own experience: "The first time it told me about one of those [conflicts]—this is like early in development—it told me three minutes before the thing. It’s like, ‘Hey, this thing is coming up. You might want to do something about it.’ I only have three minutes!" He expects the timing algorithms to improve as the model learns better context windows.

Proactive Research and Data Privacy

A key differentiator for dots is its "proactive research" feature. On a set cadence, the agent reviews what it knows about the user, including past conversations and accessible resources, to determine how it can be more helpful. It then proposes specific actions or insights. This approach creates a tension between utility and privacy, particularly given OpenAI’s broader business model.

Embiricos confirmed that dots collects data to build a profile of the user, similar to how Meta’s Muse assistant has been criticized for building detailed profiles of users and their social circles. However, OpenAI states that dots adheres to the same privacy controls as ChatGPT. Users can turn off model training to prevent their conversations from being used to improve the AI, and they can control which resources the agent accesses. Embiricos admitted that the current controls are not yet granular enough, noting that the app currently refuses read-only access to cloud drives. "What I think we need to do is basically build out super, super, super granular controls," he said, while pointing to the model’s discretion as a mitigating factor.

Scaling Agent Interactions

The broader challenge for dots, and the AI agent industry at large, is scalability. While agents are useful for individual productivity, they pose infrastructure challenges when they begin interacting with external systems. Embiricos noted that businesses are not always equipped to handle agent-driven requests, and many consumers are resistant to dealing with non-human entities. "I personally never intend to have a meeting with someone else’s agent," Embiricos said, adding that he still relies on a human assistant for corporate interactions because partners prefer human contact.

Despite these hurdles, OpenAI is exploring deeper integrations with the financial ecosystem. Embiricos revealed that the company is in "deep conversations with payment processors" to enable agents to spend money on behalf of users. This would be a potentially lucrative first for OpenAI, but it raises significant consumer trust issues. Granting an assistant access to an email inbox is already a major concession; giving it access to a bank account is a much larger ask that OpenAI is approaching cautiously.

What It Means for Developers

For developers and power users, dots represents a shift from tooling to delegation. The $100/month price point positions it as a premium productivity layer rather than a general-purpose chatbot. The integration with existing desktop and mobile apps means no new client is required, but the reliance on background processes means users must be comfortable with an AI that is constantly reading their digital life. The lack of granular permission controls is a notable gap for enterprise users who may have strict data governance policies. The proactive nature of the assistant means that workflow automation will be less about writing scripts and more about curating the context the agent has access to.

What to Watch

• Granular Controls: OpenAI has acknowledged the need for more detailed permission settings. Watch for updates that allow specific resource-level access controls.
• Payment Integration: The rollout of agent-driven payments will be a major test of consumer trust and infrastructure readiness.
• Timing Accuracy: The current lag in proactive alerts is a key usability metric. Improvements here will determine if dots is a useful assistant or a noisy notification system.
• Competitor Response: Meta’s Muse and other startups like Hark are competing in this space. Hark, in particular, is marketing its lack of data collection for advertising as a key differentiator, which may pressure OpenAI to clarify its data usage policies further.