OpenAI's Brockman: staff resent ChatGPT reaching out on a coworker's behalf
Greg Brockman, OpenAI's president and co-founder, described a pattern he has observed inside the company. Many OpenAI employees connect their personal ChatGPT accounts to the company's Slack, letting the assistant reach out to coworkers on their behalf. Brockman says people consistently dislike it when a colleague's ChatGPT contacts them asking for help with a task, even in cases where they would have been happy to do that same work if the coworker had asked them directly. He frames the reaction as telling: it shows how much people care about human relationships and about helping each other, and that they want AI to give time back to them, or to enhance the time they spend together, rather than become a layer that separates people. The remark was published as a quote by developer and commentator Simon Willison on his blog. No specific example of what a coworker's ChatGPT might ask for help with is given, and no data or survey backs the claim that this reaction is widespread.
Key facts
- Many OpenAI employees connect their personal ChatGPT accounts to the company's Slack.
- Coworkers dislike it when a colleague's ChatGPT contacts them asking for help with a task, even if they would have agreed had the colleague asked directly.
- Greg Brockman, OpenAI's president and co-founder, says the reaction shows how much people value human relationships and helping each other directly.
- Brockman says people want AI to give time back or enhance time spent together, not become a layer separating people.
- The statement is quoted by Simon Willison on his blog, with no specific example of the task involved and no survey data behind it.
Why it matters
The observation comes from inside a company that builds the tool in question, which gives it some weight as a real-world signal rather than outside speculation. It captures a tension that grows as personal AI assistants get wired into everyday work tools: an assistant acting on someone's behalf can save that person time while costing the recipient goodwill, because the message no longer carries the social weight of a direct human ask.
Who it affects
OpenAI staff who use ChatGPT integrated with Slack, and by extension any team or company considering letting personal AI assistants message coworkers automatically. It is also relevant to anyone designing agentic workflows that involve one person's assistant contacting another person rather than a system.
How to use it
Brockman's point suggests a practical norm for teams adopting AI assistants at work: use them to save your own time on tasks, but be cautious about letting them initiate requests to colleagues on your behalf, since the same request lands worse coming from a bot than from you directly. Where a human ask would work, sending an assistant instead can cost more goodwill than it saves in effort.
How solid is it
This is a first person account from Greg Brockman, OpenAI's president and co-founder, published as a quote on Simon Willison's blog. It rests on Brockman's own observation inside the company; no survey, no employee count, and no specific example of the kind of request involved is provided to support the generalization that people 'really don't like' this pattern.
Risks and caveats
No date or venue is given for when Brockman made the remark beyond the blog post's own posting, no task or scenario is named, and no figures back how widespread the behavior or the reaction actually is inside OpenAI. It should be read as an anecdotal observation and opinion, not a measured finding.
“people really don't like when a coworker's chatgpt contacts them asking for help with a task, even when they'd be perfectly happy doing that same work if asked by that coworker.”
— Greg Brockman, President and Co-Founder, OpenAI