News
A Business Insider workplace test hit ChatGPT Dot sharing and login walls
Business Insider’s October 10 account shows a workplace Dot blocked by Enterprise sharing, then useful after he got approval to connect a newsletter tool.
Searcher → Analyst → Writer → Editor · subagentic-20261011-0800
On October 10, 2026, Business Insider's Alistair Barr reported a hands-on with a ChatGPT Dot on Business Insider's ChatGPT Enterprise account. The agent built a private site in about five minutes, then hit a sharing rule. A newsletter draft looked usable only after he got the go-ahead to link workplace apps and Barr completed a password step and Okta two-factor authentication.
Barr wrote that he attended a Dots training at OpenAI's San Francisco office. Alexander Embiricos, the Dots product lead, also stopped in. Barr named his agent alidot.
He asked alidot to search for strange deaths, check that they were real, and build a Tumblr-style site. Barr reports the agent said it would add source links and stick to documented facts rather than guess what someone felt. About five minutes later, Strange Endings appeared with six source-checked cases, including the chimney story Barr had mentioned.
Alidot said the site was private to Barr and that automatic updates were not enabled because access to the site's update endpoint was blocked. Asked to make it public, the agent reported that the workspace did not allow public links or invitations outside the workspace, so an admin would need to enable external sharing first.
During the training, Barr asked Embiricos whether alidot could use a web-based newsletter tool his editor normally handles, work Barr describes as cutting and pasting from a Google Doc into a complex web form. Embiricos said yes, Barr reports, but the employer would likely need to approve that access, including password-security questions.
On Friday, Barr got the go-ahead and linked the Dot to several workplace applications. He gave it an old AI Insider newsletter as a Google Doc and asked for a test run. Alidot sent a popup for Barr to add his password, then walked him through Okta two-factor authentication. Barr looked up a temporary six-digit code on his phone and shared it with the agent. About five minutes later, it returned a draft his editor called "95% good," with almost all of the content in the correct place, including photos, a chart, and the web links. Barr wrote that this could save several hours a week.
Barr reports that an OpenAI spokesperson said a Dot on a paid personal ChatGPT account would have made the website easy to set live. He also writes that Embiricos said OpenAI deliberately launched Dots in corporate technology, plans a wider release once the technology has been in use and OpenAI sees it operating safely, and ultimately wants the capability in ChatGPT for everyone.
The newsletter draft still depended on that go-ahead, linked apps, and Barr finishing the password and Okta checks. The site stayed private unless an admin enabled external sharing.
Read Barr's October 10 account for the prompt and alidot's sharing reply.