OpenAI specialist dots: what enterprise teams should know

OpenAI's dots announcement moves the business conversation beyond asking an AI a question. The more interesting question is whether an agent can take responsibility for a bounded piece of work, keep track of changes and return something a person can use.

For enterprise teams, the significant part is the preview of specialist dots. This is worth understanding now, but it is not a reason to hand an agent unrestricted access to company systems.

What OpenAI has announced

OpenAI describes dots as always-on agents powered by GPT-6 Astra, with their own cloud computers and connections to apps through plugins. Its launch announcement distinguishes personal dots from specialist dots that take on defined responsibilities inside an organisation.

The initial rollout covers eligible Pro and Business Premium users. Enterprise users can try the beta when their workspace administrator enables it. Specialist enterprise dots are beginning as focused pilots, with OpenAI engineering teams working directly with organisations. An ordinary ChatGPT subscription should not be treated as automatic access to that specialist programme.

The specialist preview describes separate identities, credentials and access to company systems. OpenAI cites internal testing across procurement, invoice processing, email marketing, customer support and commercial contracting. Those examples are OpenAI's, not Blue Canvas case studies.

Buy a better workflow, not an impressive demonstration

Our recommendation is to start with a task whose inputs, output and owner are already understood. A weekly supplier-status pack is a stronger starting point than an instruction to improve procurement. A draft exception report is easier to evaluate than a mandate to manage finance.

Write a one-page role description before choosing connections. Include the business purpose, permitted sources, expected output, frequency, reviewer and actions that require approval. Name the person who can stop the pilot. If nobody owns the current process, fix that before giving it to an agent.

Three sensible enterprise pilot designs

• Supplier briefing: assemble changes from an approved set of supplier records and identify missing information. A buyer reviews the pack; the agent does not accept commercial terms.
• Sales preparation: bring together account notes and approved product information before a meeting. A salesperson checks accuracy and approves any external proposal.
• Operational exceptions: compare a daily report with agreed rules and prepare an investigation queue. The process owner decides what to change in the underlying system.

These are proposed designs to validate against actual access, integrations and product behaviour. They are not a list of plug-and-play features.

Separate access from authority

An agent being able to open a system does not mean it should be allowed to make every change a human user could make. Start with the smallest useful dataset and read-only access where possible. Add draft preparation next. Consider writes only after you can demonstrate what happens when input is missing, contradictory or malicious.

OpenAI describes Custom Rules, action review and mandatory safeguards in its safety explanation. Those protections do not replace your organisation's own access review, supplier assessment or incident process.

Measure completed work, including review time

Record how long the task takes today, how often it needs correction and what failure would cost. During a pilot, measure the combined time spent preparing, reviewing and fixing the agent's output. A fast draft that takes longer to check is not an efficiency gain.

Agree a stopping rule before launch. Pause if the agent exposes the wrong records, makes an unauthorised change, repeatedly omits important exceptions or cannot show where a material claim came from. Scale only when the reviewer can explain why the output is dependable for that particular task.

For more specific starting points, read our guides to dots for lawyers and legal teams and dots for finance teams.

Choose a useful first pilot

Blue Canvas can help you map a workflow, define its approval points and decide what to test before connecting business systems. Start with an AI audit or explore our AI implementation and automation support.

Book a free 15-minute consultation. Bring one recurring task, the systems involved and the decision you would never want an agent making alone.

Sources and scope

Based on OpenAI's dots announcement and safety, security and privacy explanation, published on 29 September 2026. Workflow designs below are Blue Canvas recommendations, not tested client results or promises of product capability. Blue Canvas is an independent consultancy; this article does not imply OpenAI partnership or access to its specialist enterprise pilots.