A real-world test of OpenAI’s Dots shows AI agents completing complex office tasks in minutes, while company security permissions determine how much work they can actually take over.
WHAT’S HAPPENING
Artificial intelligence is moving beyond answering questions and generating content into something businesses have wanted for decades: technology that can independently carry out everyday office work.
OpenAI’s new Dots product is demonstrating both the potential of autonomous AI assistants and the practical obstacles that still prevent widespread workplace adoption.
In October 2026, Business Insider technology journalist Alistair Barr attended an OpenAI training session in San Francisco, where participants were introduced to Dots, an AI agent system designed to handle multistep assignments across digital applications.
Barr created an agent and gave it two different tasks.
The first involved researching unusual historical deaths, verifying sources, and building a website displaying the findings.
The agent completed a private website containing six documented cases in approximately five minutes.
However, when Barr asked it to make the website publicly accessible, the system encountered a corporate security restriction. Business Insider’s enterprise workspace prohibited external sharing without administrator approval.
The second assignment involved a routine publishing task.
After obtaining workplace authorization, Barr asked his AI agent to transfer an existing newsletter from Google Docs into the company’s publishing platform.
The agent navigated the application, completed authentication with assistance, and assembled the newsletter, including text, images, charts, and links.
The task took approximately five minutes.
Barr’s editor described the finished draft as approximately 95% correct.
That assessment represents one editor’s judgment of a single test, not an independently measured accuracy rate.
Nevertheless, the demonstration illustrates how AI agents are beginning to perform real business operations rather than simply offer suggestions.
WHY IT MATTERS
Businesses spend considerable time on administrative work that requires employees to move information between applications, complete forms, prepare reports, and maintain digital records.
These activities are necessary but frequently repetitive.
Traditional automation often requires custom software integrations, specialized programming, or changes to existing business systems.
AI agents introduce another approach.
Instead of building a separate integration for every task, an AI assistant may be able to operate existing applications through their interfaces, following instructions and completing defined workflows.
For example, a small business might use an AI agent to transfer information between customer records, prepare recurring reports, organize documents, or draft routine communications.
The productivity potential is considerable, particularly for organizations without dedicated administrative teams.
But the Business Insider experiment exposes an important limitation.
An AI agent’s technical ability to perform a task does not mean it automatically has permission to perform that task.
Corporate security policies, passwords, authentication requirements, application restrictions, and employee access levels still determine what an agent can accomplish.
Those controls are not necessarily obstacles that should be removed.
They protect confidential information, prevent unauthorized activity, and establish responsibility for actions performed within business systems.
The challenge is enabling useful automation without weakening those protections.
WHO BENEFITS
Small and Mid-Sized Businesses — Could automate repetitive administrative functions without immediately hiring additional employees or commissioning expensive custom software.
Office Workers and Administrative Teams — May spend less time transferring information, preparing documents, and completing repetitive digital tasks.
AI Technology Providers — Gain opportunities to develop autonomous assistants capable of operating across business applications and completing assigned workflows.
IT Departments and Security Providers — Become increasingly important in determining how AI agents receive access, authenticate, and operate within approved systems.
Enterprise Software Companies — May benefit from developing secure integrations that allow AI agents to work within established applications.
WHO LOSES
Providers of Labor-Intensive Administrative Services — May face competitive pressure as businesses automate routine processing and document preparation.
Companies With Poorly Connected Software Systems — Could struggle to adopt AI automation when applications lack reliable access methods or compatible interfaces.
Businesses Without Clear AI Access Policies — Risk unauthorized actions, data exposure, or inconsistent automation practices.
Workers Performing Highly Repetitive Digital Tasks — May see portions of their responsibilities automated as AI agents become more dependable.
WHAT HAPPENS NEXT
The next stage of workplace AI adoption will depend less on whether an agent can demonstrate an impressive task and more on whether it can perform that task reliably, repeatedly, and securely.
OpenAI has indicated that Dots will continue expanding its capabilities and integrations while maintaining organizational controls.
For businesses, the practical opportunity is to identify tasks that consume significant employee time but follow established procedures.
Document preparation, information transfers, recurring reporting, and content publishing are potential starting points.
Companies should begin with controlled assignments, verify the results, and measure actual time savings before expanding automation.
The Business Insider test also demonstrates why human review remains important. A draft considered 95% correct may still contain errors requiring attention before publication.
Businesses must also determine which actions require employee approval and which can safely be completed automatically.
As AI agents mature, organizations may increasingly measure productivity by completed business outcomes rather than the number of individual AI prompts employees use.
The technology could also change how companies evaluate software purchases, favoring applications that AI assistants can navigate and operate securely.
The larger signal: AI is beginning to perform the digital work that employees previously handled manually. The competitive advantage may belong to businesses that successfully combine autonomous AI capabilities with secure access, reliable execution, and meaningful human oversight.
