OpenAI is adapting technology developed for its coding tools into a personal AI assistant designed to work continuously in the background, monitor users’ activities and suggest ways to handle tasks before they ask for help.
Called Dots, the assistant connects with resources such as calendars, email accounts, computers and other tools to help users organise work and respond to emerging problems. The product reflects a broader push across the AI industry to move beyond chatbots towards assistants that can take initiative and carry out tasks independently.
Alexander Embiricos, who leads OpenAI’s enterprise product team, discussed the development of Dots at TechCrunch Disrupt in October. OpenAI introduced the product at its developer event in late September 2026.
Dots is available through ChatGPT’s desktop and mobile applications to subscribers on the Pro tier and above, with plans starting at $100 per month, according to the report.
From coding assistant to personal agent
OpenAI built Dots by repackaging capabilities that were already possible through its Codex coding application into a more accessible experience for everyday users.
The company had previously adapted its agent-building work for software developers into broader productivity tools. Dots extends that approach by presenting the AI as a persistent assistant rather than a tool that requires a separate prompt for every task.
Embiricos said the goal was to make interacting with an AI agent feel more natural by giving users one continuing relationship with a single assistant.
Instead of repeatedly asking the AI what to do next, users can allow it to work in the background and identify opportunities to help.
Proactive assistance is still improving
One of Dots’ central features is its ability to monitor available information and alert users to potential issues, including calendar clashes, approaching deadlines and technical problems.
Embiricos acknowledged that the timing of these alerts has not always been effective. During early testing, his assistant once notified him about a scheduling conflict just three minutes before the event.
He expects the system to become better at anticipating users’ needs as development continues.
He also described an example involving an engineer at a drone-delivery startup whose Dot reportedly detected a service outage, analysed the problem and supplied useful information before the engineer had noticed the disruption.
Such capabilities could help workers respond to problems sooner, although the usefulness of the assistant depends on how accurately it identifies issues and when it delivers its recommendations.
Access to personal data raises privacy concerns
For Dots to provide personalised assistance, it needs access to information about users’ schedules, communications and other activities. That creates questions about how much personal information users should share with an AI system.
OpenAI’s proactive research feature periodically reviews what the assistant knows from previous conversations and the resources it has permission to access. It then looks for ways to provide additional help and presents suggestions to the user.
The approach raises concerns about how personal information may be processed by AI companies, particularly when assistants can build a detailed picture of someone’s habits, relationships and responsibilities.
OpenAI says Dots follows the privacy controls available in ChatGPT. Users can disable model training and manage which resources the assistant can access.
However, the report notes that the controls are not yet sufficiently granular for every use case. For example, the application may refuse read-only access to cloud drives, creating pressure for users to grant broader permissions.
Embiricos said OpenAI needs to develop more detailed controls. He also acknowledged that the company initially adopted conservative defaults that caused the assistant to request permission frequently, adding that those settings could be adjusted over time.
Can AI agents work across the wider economy?
Beyond privacy, the growing use of AI agents presents a practical challenge: businesses and digital services must be able to accommodate software that acts on behalf of customers.
An assistant may be able to research a purchase or identify a service, but completing transactions independently requires cooperation with websites, businesses and payment providers.
Embiricos said OpenAI was working closely with other companies as it expanded the ways its agents interact with external systems. He also disclosed that the company was in discussions with payment processors.
Allowing Dots to make payments could eventually extend its role from organising information and recommending actions to completing purchases. However, the report does not establish when such functionality might become available.
Embiricos also noted that he still uses a human assistant because OpenAI’s business partners want to interact with people directly. He said he does not intend to meet with another person’s AI agent.
OpenAI bets on a more independent assistant
Dots represents OpenAI’s effort to make AI agents easier to use by turning existing technical capabilities into a persistent personal service.
Its success will depend not only on how intelligently it handles tasks, but also on whether users trust it with sensitive information, can control its permissions and receive useful recommendations at the right time.
For now, the assistant remains an evolving product. More advanced integrations, including payments, could broaden its usefulness, but they will also increase the importance of privacy safeguards and clear limits on what an AI agent can do independently.
Source: TechCrunch