The Evolution of Always-On Artificial Intelligence
Background AI personal assistants that operate continuously—even when users aren't actively prompting them—have emerged as this year's defining approach to deep learning models. OpenAI has officially stepped into this competitive space with dots, a proactive assistant designed to handle persistent tasks across desktop and mobile environments.
From Coding Tools to Persistent Personal Assistants
The journey to dots began with OpenAI's foundational development in agentic software engineering. Speaking at TechCrunch Disrupt, Alexander Embiricos, who leads OpenAI’s enterprise product team, explained that the underlying mechanics of dots evolved directly from Codex—the company's coding tool released earlier this year—as well as ChatGPT Work.
While developers could previously replicate similar background automated workflows using Codex, doing so required significant manual effort and supervision. OpenAI recognized that converting these complex backend capabilities into an intuitive consumer product was less about creating entirely new core technology and more about reframing the user experience.
“As we were working on the product, we realized if we could frame this as, it’s an entity and it’s a single conversation, not many different conversations... it starts to become intuitive.”
— Alexander Embiricos, Enterprise Product Lead at OpenAI
Even the product's name reflects this design philosophy. A play on the negative space found inside OpenAI's corporate logo, the moniker "dots" was chosen so that the tool would feel subtle, natural, and eventually ubiquitous. (Embiricos noted lighthearted frustration at discovering very late in development that rival AI entrepreneur Elon Musk owns the dot.com URL).
Core Specifications and Ecosystem Integration
Dots Feature & Access Breakdown
- Availability: Accessible inside the ChatGPT desktop and mobile applications.
- Tier Requirements: Available exclusively to subscribers on the ChatGPT Pro tier or higher, with plans starting at $100 per month.
- Integrations: Connects directly to user calendars, email, personal computers, and surrounding workspace tools.
- Operational Model: Runs persistently without requiring active prompts, leading to higher token usage while proactively driving tasks forward.
By keeping the assistant active in the background, users can delegate continuous workflows. Embiricos shared an anecdote of celebrating when his own wife began leaving her computer running overnight specifically so her active dots agents could keep working uninterrupted through the night.
Proactive Intelligence and Real-World Performance
Unlike conventional chatbots that sit idle waiting for input, dots utilizes a continuous background feature that OpenAI calls proactive research. On a regular cadence, the assistant analyzes what it knows about the user, evaluates previous conversations, reviews connected file permissions, and autonomously researches ways to offer actionable assistance before being asked.
While testing shows that dots effectively flags calendar conflicts and impending project deadlines, early iterations revealed timing challenges. During initial development, the assistant would occasionally issue alerts just three minutes before a scheduled meeting. However, proactive capabilities have already demonstrated substantial value in enterprise scenarios.
“We had this outage, and by the time I noticed that the outage was going on, my dot had already noticed the outage was happening, had already analyzed the outage, and had come to me with useful information for how we could resolve the outage.”
— Drone Delivery Startup Engineer (Shared by OpenAI)
Navigating Privacy, Data Security, and Granular Controls
As background AI agents demand deeper access to personal files and schedules, privacy concerns inevitably arise. Competitors like Meta's assistant Muse have faced scrutiny over how personal profiles are constructed to assist users while potentially feeding corporate advertising networks. In contrast, startups like Hark explicitly market themselves on zero ad-data collection.
OpenAI clarifies that dots operates under the standard privacy framework established for ChatGPT:
- Model Training Controls: Users retain the option to disable model training, preventing OpenAI from utilizing their conversations to train underlying models.
- Resource Restrictions: Users can control which system resources and files the assistant can inspect on their machines.
- Permission Protocols: Dots initially launched with conservative default settings that frequently request explicit user permission before executing actions.
However, limitations remain. Currently, the application lacks highly granular permission settings—for instance, refusing to work under read-only access for cloud drives. Embiricos emphasized that OpenAI recognizes the need to build out significantly more detailed permission toggles over time as user trust matures.
Future Outlook: Financial Partnerships and Scaling Agents
As background AI assistants expand, questions regarding ecosystem infrastructure and human social acceptance persist. Many individuals and businesses remain hesitant to interact directly with automated AI agents. Embiricos himself noted that he relies on a human assistant because enterprise partners expect human interaction, stating he does not personally intend to hold meetings with another person's AI agent.
To overcome operational boundaries, OpenAI is focusing on deep ecosystem partnerships rather than rushing raw capabilities to market. The company confirmed it is currently engaged in extensive discussions with major payment processors. Enabling dots to handle financial transactions safely on behalf of users would mark a significant milestone for AI productivity tools—provided consumers are comfortable granting an assistant access to monetary channels alongside their email and calendars.