OpenAI launches dots as a proactive AI personal assistant

From chatbots to proactive agents

OpenAI recently introduced dots, an AI personal assistant designed to operate as a persistent entity rather than a simple chat interface. Unlike traditional models that wait for a user to type a specific command, dots is intended to work in the background of a user's digital life. The goal is to create a single, continuous conversation where the AI understands the user's ongoing context.

The company developed dots by evolving its existing agentic coding tools, which were originally built for software engineers. By repackaging this technology, OpenAI created a product that can perform complex tasks with less hand-holding from the human user. The name dots is meant to feel like a ubiquitous part of the user's workflow, rather than a separate application.

Alexander Embiricos, who leads OpenAI’s enterprise product team, explained that the focus was on making the experience feel intuitive. He noted that the transition from a chatbot to an assistant relies heavily on how the product is framed to the user. Instead of building entirely new technology, the company focused on how the existing models could feel like a helpful companion.

How the assistant works

To function effectively, dots requires access to a user's calendar, email, and various computer tools. Once connected, the agent attempts to manage work tasks proactively by thinking ahead of the user. For example, it can flag upcoming deadlines or identify potential conflicts in a person's schedule before they occur.

A core feature of the assistant is what OpenAI calls proactive research. On a regular cadence, the agent analyzes the information it has been given access to and considers previous conversations to determine how it can be more useful. It then proposes specific actions or information to the user based on these findings.

While the technology is still in development, the proactive nature is the defining characteristic of the product. Embiricos shared an example of an engineer at a drone delivery startup whose dot noticed a system outage and analyzed the problem before the human even realized it was happening. The assistant then provided the engineer with useful information on how to resolve the issue.

Privacy and data permissions

The shift toward persistent agents creates a tension between the amount of data an AI needs to be useful and the privacy of the user. Because dots must understand a user's life to be proactive, it requires significant access to personal information. OpenAI states that dots follows the same privacy controls as the standard ChatGPT platform.

Users currently have the ability to turn off model training, which prevents OpenAI from using their private conversations to improve its AI models. They can also control which specific resources the agent is allowed to access on their computer. However, these controls are not currently granular, and the app can be firm about the permissions it requires to function.

Embiricos admitted that the company has faced feedback regarding the agent's conservative defaults, which often cause it to ask for permission frequently. He noted that the team is looking to build out much more granular controls in the future. This balance between the model's intelligence and the user's desire for discretion remains a primary challenge for the product.

Scaling and the future of agents

A major question for the AI industry is whether these agents can scale to work with the broader infrastructure of the internet. While agents can successfully book hard-to-get reservations in small numbers, they may face hurdles when interacting with large-scale business systems. There is also a social hurdle, as many people and businesses still prefer interacting with humans over automated agents.

OpenAI is currently navigating these infrastructure challenges by working closely with the rest of the ecosystem. Embiricos mentioned that the company is in deep conversations with payment processors to explore how agents might eventually handle transactions. Allowing an AI to spend money would be a significant step for both the company and the consumers using the tool.

For now, the assistant remains a tool for high-level organization and research rather than a fully autonomous financial agent. While giving an AI access to an email inbox is a significant request for many users, bank accounts remain off-limits for the time being. The company is moving deliberately to ensure that as agents gain more capabilities, they do so in collaboration with existing systems.

Availability and requirements

Dots is not available to all ChatGPT users; it is specifically designed for the Pro tier and higher. Subscribers can access the assistant through both the desktop and mobile applications. The starting price for these plans is $100 a month, positioning it as a premium tool for power users.

The product is intended to be a persistent companion that stays active even when the user is not actively interacting with it. This persistent nature is what allows it to perform the background research and proactive notifications that distinguish it from standard chat models. It aims to reduce the amount of manual prompting required to keep a user's schedule organized.

As the product matures, OpenAI expects the timing of its proactive notifications to become more precise. Early versions may flag a conflict only a few minutes before it happens, but the goal is for the assistant to provide timely alerts. By refining these interactions, OpenAI hopes to make dots a ubiquitous part of the daily workflow for its professional subscribers.

Try it today

If you are a ChatGPT Pro subscriber, you can check your app settings to see if dots is available for your account and explore the permissions you can grant it for your calendar and email.

Watch

How OpenAI plans to implement more granular privacy controls for dots.

Whether OpenAI will successfully secure partnerships with payment processors to allow agents to spend money.

How the timing of proactive notifications improves as the model matures.