Meet OpenAI Dots: The AI Agent That Works After ChatGPT Closes

OpenAI Dots are persistent AI agents designed to continue delegated work after a ChatGPT conversation ends, using cloud-based tools, connected information and ongoing context. For Sydney and NSW property operators, their potential lies in maintaining continuity between changing project instructions, documentation and site coordination. However, autonomous assistance does not replace verified site information, statutory approvals or human responsibility for commercial and construction decisions.

OpenAI Has Introduced an AI Agent That Does Not Wait for Another Prompt

Most people have become accustomed to a particular relationship with artificial intelligence: open a conversation, provide instructions, receive an answer and return when another task requires attention.

OpenAI's newly introduced Dots challenge that operating model.

Announced in September 2026, Dots are described as always-on AI agents capable of maintaining responsibility for delegated work between conversations.

Rather than treating every interaction as an isolated request, a Dot can retain relevant context, undertake ongoing tasks, investigate developments and return to its user when further information or a decision is required.

According to OpenAI's official Dots documentation, the system is powered by GPT-6 Astra and operates through its own cloud-based computer and browser.

That distinction matters.

An assigned cloud task can continue even when the user's computer is switched off. Work requiring access to a connected local computer, however, depends on that computer remaining online with the appropriate ChatGPT application running.

Dots can also use permitted application connections and communicate through supported environments, including ChatGPT, Slack and Microsoft Teams.

It is an extension of AI beyond the conversation itself. The commercial question is what happens when that persistent capability encounters an organisation whose information is constantly changing.

The Real Business Problem Is That Project Information Never Stands Still

Consider a Sydney property operator coordinating an apartment renovation involving existing flooring removal, concrete preparation and installation of a replacement finish.

The initial scope might be relatively straightforward.

A contractor has an approved quotation, the project coordinator has scheduled the removal team, the owner has selected a flooring product and the strata manager has received the proposed works documentation.

The difficulty begins when something changes.

The client selects a different flooring system. A supplier advises that the nominated product is unavailable. A contractor identifies an unexpected substrate condition. The building manager introduces a revised access restriction.

Each development may reach a different person through a different communication channel.

The project manager might understand the latest client request, while the installer continues referencing an earlier scope. The supplier could be discussing a substitute product that has not been assessed against the strata acoustic requirements. Meanwhile, an installation date remains visible in the project calendar.

No single instruction is necessarily incorrect in isolation. The problem is that the combined project record is no longer consistent.

This creates a different opportunity for persistent AI agents. Rather than simply answering another question, an assigned Dot could help identify changes affecting an ongoing project, compare authorised information and prepare the unresolved dependencies for review.

The objective is not to allow AI to manage a renovation independently. It is to prevent important decisions from becoming disconnected as information moves between people, applications and project stages.

What Makes Dots Different From Conventional ChatGPT Workflows?

Dots introduce a continuing agent relationship rather than requiring every responsibility to be restarted through a new conversation.

OpenAI describes several capabilities that distinguish the system from conventional conversational assistance.

Persistent work

• How OpenAI Dots operate: Delegated work can continue between conversations.
• Potential business relevance: Longer-running responsibilities need not depend on an employee keeping a chat session open.

Cloud execution

• How OpenAI Dots operate: A Dot has access to its own cloud computer and browser.
• Potential business relevance: Research, document preparation and supported cloud tasks can continue independently of the user's device.

Ongoing context

• How OpenAI Dots operate: Relevant conversations, memory and saved working notes can inform subsequent activity.
• Potential business relevance: The agent can maintain continuity as project requirements and priorities develop.

Connected applications

• How OpenAI Dots operate: Supported applications can be used within their existing permissions.
• Potential business relevance: Information held across authorised workplace systems can contribute to the assigned responsibility.

Proactive follow-up

• How OpenAI Dots operate: Dots can determine when to resume work and request human input where required.
• Potential business relevance: Important developments can be surfaced without requiring the same manual prompt every time.

Human intervention

• How OpenAI Dots operate: Users can review activity, redirect work and manage permissions and approvals.
• Potential business relevance: Operational and commercial decisions can remain with accountable employees.

There is an important qualification: connecting an application does not automatically authorise unrestricted monitoring or execution.

The user must define the agent's responsibilities and provide access to the relevant information. Actions affecting connected accounts or sharing information are also subject to OpenAI's applicable permission and approval mechanisms.

Dots are being introduced progressively across eligible ChatGPT Pro, Business Premium and Enterprise plans. Actual access depends on account eligibility, regional availability and relevant workspace administration settings.

A Sydney Project Example: When One Flooring Change Affects Four Other Decisions

The practical implications become clearer when Dots are considered as a project-continuity layer rather than an automated personal assistant.

Consider the following illustrative scenario. It describes a potential configured workflow, not an existing Elyment deployment or a guaranteed out-of-the-box Dots integration.

The Original Project

A strata apartment in Sydney is scheduled for carpet removal, substrate preparation and installation of a new hard flooring system.

The project file contains the approved scope, selected product details, proposed programme and supporting acoustic documentation.

The Change

The client requests a different flooring product after the initial documentation has been prepared.

A persistent agent assigned to support project documentation, and given access to the relevant records, could investigate whether that change introduces inconsistencies elsewhere in the project.

For example:

• Does the available acoustic documentation refer to the original product or the proposed replacement system?
• Has the supplier confirmed availability of the replacement material?
• Will the new flooring build-up alter the assumed finished floor height?
• Does the existing installation programme remain achievable?
• Has the contractor received an approved scope revision?

Instead of making those decisions independently, the Dot could prepare a consolidated exception report and draft the required clarification requests.

The project coordinator would then assess the proposed changes with the appropriate contractor, supplier, building manager or technical adviser.

Why This Matters Under NSW Strata Requirements

The distinction between an administrative product change and a technically significant project variation is particularly important in NSW strata buildings.

The NSW Government's strata renovation guidance identifies the installation or replacement of hard flooring, including carpet replacement, as a minor renovation generally requiring approval under the applicable strata arrangements.

The guidance also identifies acoustic certification among the information required for flooring applications. Individual scheme by-laws and the nature of the work must still be considered.

An AI agent could help identify that an approval document refers to an earlier product. It cannot determine that a replacement flooring system automatically satisfies the building's acoustic requirements.

That remains a matter for appropriate documentation, technical assessment and authorised decision-making.

For further context, Elyment has examined the relationship between flooring works and ownership boundaries in Strata Says It Is Common Property: What That Means Before You Renovate the Floor.

The Most Useful AI Output May Be a Record of What Has Changed

Many organisations already possess more project information than their employees can comfortably review during a busy working day.

The issue is frequently not missing data. It is uncertainty about which information remains current.

This is where a persistent AI agent could support a more disciplined approach to project documentation.

For an assigned project, an effective working record could distinguish between four information states:

1. Confirmed: Information supported by the current approved documentation.
2. Changed: A newer instruction, correspondence item or document has introduced different information.
3. Pending verification: The change affects another requirement that has not yet been confirmed.
4. Approved for action: The authorised person has reviewed the evidence and released the next operational step.

This is a proposed business workflow, not a claim that Dots automatically provide a construction-specific document-control system.

The distinction is commercially important.

An unanswered client email does not necessarily represent an approved variation. A supplier's alternative product is not automatically a suitable substitute. An updated calendar entry does not establish site readiness.

Persistent AI becomes more useful when its instructions require it to preserve those distinctions rather than treat every newly discovered piece of information as authority to act.

Where Autonomous AI Must Stop Before Physical Work Begins

Digital continuity and physical readiness are not interchangeable.

A Dot may have access to the latest approved drawings, contractor correspondence and floor preparation notes. It does not follow that the actual site remains in the same condition described by those documents.

For example, flooring removal can expose existing adhesive, unexpected layers, damaged screed, moisture concerns or substrate irregularities that were not visible during the initial inspection.

Those discoveries may affect grinding requirements, floor levelling quantities, disposal allowances, installation methods and the final project programme.

Elyment's existing guide to sequencing floor removal, grinding, levelling and installation on busy Sydney sites explains why each physical stage depends on the condition left by the preceding activity.

An always-on agent could help reconcile the resulting documentation. It should not independently certify the condition of the exposed substrate or authorise an unassessed change in construction methodology.

The practical division of responsibility is:

Scope changes

• Potential agent responsibility: Identify affected documents and draft a revision summary.
• Required human responsibility: Approve the revised scope and commercial consequences.

Substrate discoveries

• Potential agent responsibility: Organise photographs, observations and outstanding questions.
• Required human responsibility: Assess site conditions and determine the appropriate treatment.

Supplier changes

• Potential agent responsibility: Compare available specifications and delivery information.
• Required human responsibility: Approve substitutions and relevant technical requirements.

Strata requirements

• Potential agent responsibility: Identify missing documentation or conflicting information.
• Required human responsibility: Obtain and verify the necessary approval.

Installation scheduling

• Potential agent responsibility: Prepare revised programme options.
• Required human responsibility: Confirm access, labour availability and physical readiness.

Client communication

• Potential agent responsibility: Prepare an evidence-based update for review.
• Required human responsibility: Authorise commitments affecting scope, price or completion.

The operating principle is that an agent can keep information moving without being given independent authority over the physical consequences of that information.

Security and Privacy Become More Important When the Agent Has Ongoing Context

Persistent agents raise a further operational issue: information access is no longer limited to a short, isolated conversation.

A Dot may work with relevant context over time and use authorised connected applications to support an ongoing responsibility.

For NSW property operators, this could involve client correspondence, quotations, contractual information, supplier records, project documentation and potentially personal information.

Organisations must consider whether the information being connected is appropriate for the intended task.

The Office of the Australian Information Commissioner's guidance on commercially available AI products emphasises privacy due diligence, transparency, appropriate handling of personal information and human oversight.

Organisations covered by the Privacy Act 1988 must consider their applicable Australian Privacy Principles obligations.

The Australian Signals Directorate's Careful Adoption of Agentic AI Services guidance also recommends incremental deployment, controlled privileges, visibility and human oversight.

OpenAI states that Dots incorporate existing ChatGPT permissions, built-in safeguards, automatic review of applicable actions and optional Custom Rules.

However, product-level safeguards do not remove the organisation's responsibility for its own information governance.

A practical starting point for a Sydney business would be to assign a Dot one clearly defined responsibility, connect only the necessary information sources and require review before any external or commercially significant action.

How Sydney Businesses Could Test Dots Without Rebuilding Their Operations

The first useful deployment does not need to involve complete business automation.

A property operator, renovation coordinator or commercial project manager could begin with one active responsibility: maintaining consistency across the documentation for a selected project.

A controlled pilot could follow five stages.

1. Select one project. Choose an active renovation or fit-out with a defined scope, an accountable coordinator and accessible project records.
2. Establish the approved information. Identify the current quotation, scope, project programme, relevant approvals and documented assumptions.
3. Define the agent's responsibility. Instruct it to identify relevant changes, investigate inconsistencies and prepare proposed follow-up actions.
4. Keep project authority with employees. Require human approval before changes to bookings, external communications, procurement or contractual commitments.
5. Review the results. Evaluate whether the agent identified useful inconsistencies, introduced incorrect assumptions or reduced repetitive document-checking work.

The appropriate performance measures are not simply the number of messages generated or tasks completed.

A more useful evaluation would examine how reliably the agent identifies document conflicts, attributes information to the correct source, distinguishes pending decisions from approvals and prepares work that an employee can confidently review.

Subscription allowances, connected application requirements, staff supervision and workflow setup should also be considered before assuming that persistent operation will reduce costs.

The Broader Shift: AI Is Becoming Part of Operational Continuity

The introduction of Dots reflects a change in the relationship between employees and workplace AI.

The earlier model concentrated heavily on producing an immediate answer. The emerging model allows responsibility for a continuing piece of work to remain active between interactions.

For the Sydney property and construction sector, that development is relevant because project information rarely arrives in a predictable order.

Client preferences change. Suppliers revise delivery information. Contractors discover new site conditions. Building managers introduce restrictions. Approvals may remain outstanding while the programme continues to develop.

Persistent AI could help organisations maintain a clearer record of how those developments affect ongoing work.

It does not eliminate the need for competent contractors, technical advisers, project managers or properly documented approvals.

The distinction between a conversation that has ended and a responsibility that remains active is nevertheless significant.

OpenAI Dots demonstrate how an AI agent can continue working after a user steps away. For property operators, the real commercial value will depend on whether that continuity produces more reliable information, clearer handover decisions and fewer unresolved contradictions before instructions reach a live site.

Elyment | Project & Operational Review

Keep Project Decisions Connected to Delivery

Review renovation workflows, documentation, approval requirements and operational dependencies before disconnected information affects your next project stage.

Request a Project Review →

Sources and Further Reading

OpenAI Product Documentation

• OpenAI: Meet Dots, Official Product Documentation
• OpenAI: Introducing Dots

NSW Government, Privacy and Cybersecurity Resources

• NSW Government: Strata Renovation Rules
• OAIC: Privacy and Commercially Available AI Products
• Australian Cyber Security Centre: Careful Adoption of Agentic AI Services

Related Elyment Articles

• Strata Says It Is Common Property: What That Means Before You Renovate the Floor
• How Should Busy Sydney Sites Sequence Removal, Grinding, Levelling and Installation to Avoid Trade Clashes?