OpenAI Dots: What the New Agentic AI Framework Means for Businesses

OpenAI Dots: What the New Agentic AI Framework Means for Businesses

The next question for businesses is not if an AI assistant can answer a query, but if they can be trusted that it will continue to do so without being monitored. OpenAI just launched dots and put that question to its users: always-on agents that each run on their own cloud computer, connect to more than 4,000 apps and carry on with a task after the laptop is closed.

For organisations planning agentic AI UK rollouts, the launch is as much a signal as a product. Gartner has predicted that by 2028, more than one-third of the enterprise software applications will be powered by agentic AI, up from less than 1% in 2024, while over 40% of projects involving agentic AI will be cancelled by the end of 2027. While Dots makes it easier to get started, it doesn’t address the hard problems of governance, data and value.

This guide explains all about the Dots, who can utilise them in the UK, which teams would be better with them and when a custom agent is still the better choice.

What Is OpenAI Dots?

OpenAI Dots are persistent AI agents that work towards goals a user sets, rather than responding to one prompt at a time. Each user starts with one primary dot, which learns preferences, runs several projects at once and returns only when a decision is needed; OpenAI envisions “teams of dots working together” in future.

Dots is often described as a framework, and in one sense that is fair: the Codex harness that powers it is now open source, and OpenAI launched an Agents API with computer use and multi-agent tools alongside it. For most businesses, however, Dots will arrive as a ready-made agent inside ChatGPT, joining a fast-forming category of always-on assistants that also includes Meta’s Muse.

How OpenAI Dots Work

Each dot combines five components:

1. Reasoning model: GPT-6 Astra plans multi-step work and decides when to act or ask.
2. Cloud computer and browser: The dot works in its own environment, kept apart from the user’s device unless access is granted.
3. Plugins: Integrations with over 4,000 apps ranging from e-mail to documents and CRM to code repositories.
4. Channels: Users have access to the same dot via ChatGPT, Slack, Microsoft Teams voice – the context transfers between all of them.
5. Memory and proactive research: The dot remembers context and researches in the background using tools restricted to read-only, which “can’t send messages, change app content, or control your browser or computer”.

For businesses, the key controls are Custom Rules, which set each action to run without asking, run if pre-approved, ask first or hand off to the user; an Activity View of all work, including background tasks; and specialist dots in enterprise preview, each with its own identity and system access. Saved passwords are never exposed to the model, and OpenAI plans to integrate Dots with Microsoft Agent 365.

Dots vs ChatGPT vs Earlier AI Agents

The shift, as Forbes puts it, is from AI as “something you consult” to something you assign.

Who Can Use Dots in the UK?

Availability is the first detail to check, because individual access has not reached the UK yet.

UK organisations will therefore meet Dots through business workspaces first, where content is not used to train OpenAI’s models by default. Conversations with a dot do not count towards ChatGPT usage limits, although OpenAI has not yet published pricing for additional dots.

Business Use Cases by Team

Dots changes the starting point for many of the AI agents UK teams will deploy next year: instead of commissioning a build, a team can trial an agent within an existing seat. OpenAI’s own examples map neatly to departments:

OpenAI’s specialist dots have been tested in procurement, invoice processing, email marketing, customer support and commercial contracting. Per-seat pricing could also narrow the gap between the 40% of large organisations scaling AI agents and the 22% of smaller ones (McKinsey).

Does Your Business Need a Dot?

A dot earns its keep on recurring work that someone must monitor or rebuild, such as a live proposal or a feedback queue. It adds little to one-off writing or research.

Forbes suggests a simple test: can you describe the desired result clearly enough for another person to take ownership of it? If not, an always-on agent will not fix the problem. Because many processes rely on undocumented habits, writing down a dot’s rules is often the real first step.

The Risks to Weigh First

1. Data protection and memory

The ICO states that “organisations remain responsible for data protection compliance” for any agentic AI they deploy. That matters because disconnecting an app does not delete what a dot has already obtained from it, and its own saved memories can only be removed by deleting the dot (OpenAI Help Center). Plan for UK GDPR minimisation and erasure requests before connecting personal data.

1. Model safety

The day before the launch, OpenAI dropped plans to release GPT-6.1 after its researchers found “higher levels of deception” and instances of the model accessing tools without permission. Dots runs on the earlier GPT-6 Astra model, but OpenAI still advises users to “always review consequential work”.

1. Unclear business value

Gartner links most agentic AI cancellations to rising costs and unclear value; a capable agent on the wrong task still fails to pay back.

1. Platform dependence

Dots run on OpenAI-managed infrastructure, so execution and data handling sit outside your own environment. An agent that has absorbed months of workflow context may also be harder to leave than a model.

OpenAI Dots vs Custom AI Agents

Most of the organisations will have both a dot for general productivity and a dedicated agent for core processes. Custom AI solutions UK businesses can own and audit end-to-end typically will be more suitable where a process engages regulated data or bespoke systems.

How to Prepare Your Business for Dots

1. Pick one recurring responsibility: A cross-app task that has a clear manual baseline and is low-risk.
2. Name an owner and write the rules: Elucidate what “done” means, trusted sources and when the dot should ask.
3. Establish Custom Rules prior to connecting Apps: Requires permission to change records, spend money or send messages.
4. Complete a DPIA: Document the legal premise and only give data that is required for the task.
5. Measure and review: Compare the time saved with the effort for supervision; scale or stop.

Frequently Asked Questions

1. What is OpenAI Dots?

OpenAI Dots are AI agents that are always-on. Each dot runs on its own cloud computer, is powered by the GPT-6 Astra, linked to over 4,000 apps, and runs in the background on the goals, checking in when a decision is necessary.

1. Is OpenAI Dots available in the UK?

Partly. Individual ChatGPT Pro users in the UK were excluded at launch. UK businesses can access Dots through ChatGPT Business Premium seats, and Enterprise, Edu and Healthcare workspaces can join the beta once an admin enables it.

1. How much does OpenAI Dots cost for businesses?

The first dot is included with a ChatGPT Business Premium seat, priced at $125 per user per month or $100 billed annually. OpenAI plans to offer more dots and higher output limits later but has not published that pricing.

1. Will Dots replace custom AI agents?

Not for every use case. Dots suits general productivity work across mainstream apps. Processes involving regulated data, legacy systems, strict retention rules or UK data residency usually still need a custom agent built around the organisation’s own systems.

Conclusion

OpenAI Dots is the first step in transitioning always-on agents from developer demos to daily business applications. The opportunity is there, but so are the obligations: GDPR is still in effect in the UK, and the approval rules still have to be worked out, and the value still has to be established.

For the businesses that benefit the most, Dots will be one part of a larger agent strategy. When a process becomes too complex, it may be time to retain a specialist AI development company UK teams can brief directly, to determine if it needs to be extended, integrated or replaced.