OpenAI’s Dots Can Keep Working After You Leave ChatGPT. That Changes What Using AI Means

OpenAI’s Dots Can Keep Working Without You. That Changes How We Use AI

OpenAI’s Dots are persistent AI agents powered by GPT-6 Astra that get their own cloud computer and can work across connected services, including ChatGPT, Slack and Microsoft Teams. Unlike a normal chatbot session, a Dot can continue managing an assigned goal in the background, although consequential actions can still require human approval.

OpenAI has launched Dots, a new class of always-on AI agents designed to keep working on projects even when the user is no longer actively sitting inside a ChatGPT conversation.

Announced at OpenAI’s DevDay on 29 September, each Dot runs on GPT-6 Astra, receives its own cloud computer and browser, and can connect through OpenAI’s plugin ecosystem to more than 4,000 apps. Users can communicate with it through ChatGPT, Slack and Microsoft Teams.

That sounds like another AI-product launch, but the more consequential change is behavioural: the user does not necessarily have to provide every next instruction.

From asking AI questions to assigning it responsibilities

Most people currently use ChatGPT through a simple loop: ask something, receive an answer, refine the request, then leave.

Dots are designed around a different relationship. OpenAI says a Dot can take a project, continue working on it alongside other assignments and monitor new information as it arrives. Its examples include watching customer feedback and preparing software fixes, updating research when new data appears, revising sales proposals as requirements change and turning interview transcripts into content drafts.

Those are OpenAI’s examples rather than independent evidence that Dots will perform each task reliably at scale. But they illustrate the intended shift from prompting an assistant to supervising an agent responsible for an ongoing workflow.

That distinction matters for work. A worker using an AI agent may increasingly spend less time generating every individual output and more time deciding what the agent should access, what success looks like, when it should stop and which decisions require human approval.

TechView Africa has already examined why AI agents should not be allowed to perform every action unsupervised and how permissions are becoming a central AI-security problem. Dots make those questions much less theoretical.

“Always-on” does not mean unlimited authority

OpenAI describes Dots as capable of working towards a user’s goals “24/7,” but there are important limits.

When a user is away, OpenAI says Dots can perform what it calls proactive research using connected tools restricted to read-only access. Those background tools cannot independently send messages, modify app content or control the user’s computer.

Users can also define rules covering what a Dot may do independently, what requires approval and what is blocked. Some sensitive actions, including changing passwords, remain reserved for the user. OpenAI also explicitly warns that Dots can still make mistakes.

The limits matter because persistent agents have more opportunity to encounter changing information, unexpected instructions and situations their original task did not anticipate.

The first demonstrations were not flawless

There is also an early reality check. Reuters reported that several live Dots demonstrations at DevDay ran into problems when agents repeatedly failed to provide requested voice updates. OpenAI attributed the glitches to the volume of updates being rolled out around the event.

That does not establish that the underlying agent failed at long-running work, but it is a useful reminder that a product designed for persistent responsibility still has to prove its reliability outside controlled examples.

What this could change for workers

Dots are beginning to roll out to eligible Pro and Business Premium users, while Enterprise organisations can enable a beta. OpenAI is also testing specialist Dots designed around defined organisational responsibilities.

For workers, including remote professionals across Africa, the immediate implication is not that every knowledge job suddenly disappears, it is that AI supervision may become a more valuable workplace skill.

Defining tasks clearly, limiting access, checking outputs, spotting failures and deciding which actions should remain human responsibilities could become as important as knowing how to write a good prompt.

That direction is already visible in the labour market. TechView Africa’s review of AI job listings across African marketsfound employers increasingly asking for experience with agentic AI, production systems, evaluation and guardrails rather than prompt-writing alone.

Dots push that transition further. The next stage of workplace AI may be less about asking a chatbot to help with a task and more about deciding which responsibilities software can safely keep working on after you leave.

Our Recommendation

Treat Dots as an early example of a broader change from conversational AI towards persistent delegated work, not as proof that businesses can already hand entire jobs to autonomous software.

Workers should pay particular attention to AI-agent supervision, permissions, evaluation and workflow design. Employers should decide in advance which systems an agent may access, which actions require approval and who remains accountable when the agent gets something wrong.