What is a ChatGPT dot and can Scottish businesses use one?
A ChatGPT dot is an always-on AI agent that can keep working between conversations using its own cloud computer, connected apps, memory and scheduled tasks. Scottish businesses can use dots through eligible Business Premium accounts as the rollout reaches them; UK Pro accounts are excluded at launch. Start with one bounded project, limit its connections and permissions, require approval for consequential actions and review what it stores as memory.
A ChatGPT dot is designed to keep responsibility for an ongoing piece of work, not simply answer the message in front of it. It can continue between conversations, use connected apps, work on its own cloud computer, run scheduled tasks and return with progress or decisions that need your input.
That makes a dot potentially useful. It also means the sensible question is not, “What can I give it?” It is, “What is the smallest useful responsibility I can give it safely?”
What is a ChatGPT dot?
OpenAI describes a dot as an always-on agent powered by GPT-6 Astra. It has its own cloud computer and can use the apps you choose to connect. It can remember context, work through several steps, continue making progress when you are away and ask for your judgement when a decision needs it.
Examples in OpenAI's documentation include preparing and updating a presentation, turning source material into several content outputs, improving an application and monitoring a project as new information arrives.
Those examples are useful illustrations, not proof that every task should be delegated. A dot still works from the information, permissions, rules and connections it has been given. It can make mistakes, misunderstand a boundary or take an unhelpful route towards a reasonable goal.
Are ChatGPT dots available in Scotland and the UK?
Yes, for some business accounts. OpenAI says dots are available to Business Premium users across supported ChatGPT regions, which includes the UK. The rollout is gradual, so an eligible Scottish business may not see the feature immediately.
The position is different for personal Pro accounts. At launch, OpenAI excludes Pro users in the UK, European Economic Area and Switzerland. Enterprise, Edu and Healthcare workspaces can try the beta when an administrator enables it; it starts switched off.
Your first dot is included with an eligible Pro or Business Premium plan. OpenAI says deeper-work allowances are temporarily extended during the launch period and that future usage terms will follow. Treat the plan controls shown in your own account as the current position rather than building a permanent budget around an introductory allowance.
How is a dot different from ChatGPT, Work or a workspace agent?
These products overlap, but they are not the same thing.
| Product | Better suited to | Main distinction |
|---|---|---|
| ChatGPT conversation | Questions, thinking and work you direct in the current conversation | Primarily responds when you ask |
| ChatGPT Work | Longer tasks that produce a finished deliverable | Runs a substantial task while you steer and review it |
| Workspace agent | A defined workflow shared through approved workplace channels | Configured around specific tools, Skills, files and access rules |
| ChatGPT dot | Ongoing responsibility that continues across conversations and scheduled work | Has its own computer, memory and continuing relationship with its owner |
A dot is one product example of an AI agent. It does not change the underlying business principle: use an agent when the route can vary and judgement is useful. Use ordinary automation when the sequence and rules are already known.
What could a Scottish business use a dot for?
The strongest candidates are ongoing projects with clear boundaries, changing information and regular points where a person can review the work.
Maintain a live research brief
A dot could monitor named authoritative sources, update a briefing when something material changes and show what was added, removed or left uncertain. The owner should define the permitted sources and verify important claims before the briefing informs a decision.
Keep a project moving
A small team could use a dot to collect approved updates, maintain an action list, prepare a weekly summary and flag decisions that need the owner. It should not quietly expand the project scope or make commitments to clients and suppliers.
Prepare recurring content work
A dot could turn an approved interview transcript into draft notes, extracts and social copy, then carry corrections across the set. It should preserve the verified source material and leave publication under human control.
Support an internal product or coding project
A dot can create Work or Codex tasks and use an existing Codex cloud environment. That can help investigate issues, prepare changes and run tests. Source control, test evidence and human review still matter, especially when it can also reach a local computer.
These uses are not uniquely Scottish. The Scottish angle is proportionality. A small business should not copy an enterprise agent programme. Give the dot one responsibility whose value and risks the team can actually measure.
What can a dot access?
A dot begins with its own cloud computer. You can connect supported apps so it can retrieve information or take permitted actions. OpenAI also allows optional access to a local computer through the ChatGPT desktop app; this starts turned off and must be enabled.
Connected apps use the permissions available to the account that connects them. A broad account can therefore give the dot broad reach. Before connecting an app, ask:
- which records and actions does this work genuinely require?
- can the first pilot be read-only?
- can a restricted account or folder be used?
- which actions must always require approval?
- how will access be revoked if the pilot stops?
OpenAI says a dot may proactively review connected information and form memories from it. Disconnecting an app does not delete information the dot has already obtained. Its help documentation says resetting the dot deletes the dot, its conversations, saved memories and scheduled tasks.
That is a material governance point. Do not connect customer, staff or commercially sensitive systems until the business has decided what the dot may read, retain, share and act upon.
What controls does a dot provide?
OpenAI documents custom rules for actions involving sharing, purchasing and access. Depending on the action, the owner can set the dot to:
- take action without asking;
- act when the request has been explicitly pre-approved;
- ask before acting; or
- hand the action back to the owner.
The product also lets the owner inspect the dot's computer, review in-progress and scheduled work, pause it, revoke local-computer access or reset it.
OpenAI explicitly warns that a dot can make mistakes even when following custom rules. Product controls are useful, but they are not a substitute for limited permissions, logs, testing, incident handling and a manual fallback.
How should a business test its first dot?
Choose one project that is useful enough to measure but safe enough to run with restricted access.
- Define one responsibility. Write down the outcome, the approved sources and what remains outside scope.
- Limit the connections. Start with the minimum files and apps required, preferably read-only.
- Set decision boundaries. List what the dot may do independently and what needs approval.
- Use realistic cases. Include missing information, conflicting instructions, an outdated source and a failed connection.
- Review memory. Check what context it carries forward and whether that remains appropriate.
- Measure the complete result. Record useful work completed, corrections, review time, delays and any additional usage cost.
- Prepare the stop route. Know how to pause it, revoke access and continue the work manually.
Run this way for long enough to see ordinary work and awkward exceptions. One polished demonstration is not evidence that an always-on agent should receive broader access.
Should your business use a ChatGPT dot?
A dot is worth evaluating when a continuing project repeatedly needs research, coordination, drafting or technical work, and the business can define clear approval boundaries around it.
It is not the right starting point for a process nobody understands, a simple fixed reminder, an exact calculation or a high-impact workflow that the organisation cannot monitor. In those cases, improve the process first or use ordinary AI automation with a narrower AI-assisted step.
My view is straightforward: the value is not that the agent is always on. The value is that it can keep one worthwhile responsibility moving without quietly acquiring responsibility for everything else.