OpenAI's Dot Agent Sent 300 Emails to Move One Office

OpenAI's Dot agent sent roughly 300 emails in a few days while coordinating a startup's office move — and its owner shut it down after the movers complained. Robert Farlow, the 27-year-old founder of the AI startup Plato, gave Dot the job of managing his company's relocation in San Francisco. The agent triple- and quadruple-emailed movers, insurance workers and two sets of building managers, sometimes simultaneously.

Farlow's account, told in an as-told-to essay, is an early field report on what happens when autonomous agents run unsupervised on real business tasks. OpenAI did not respond to a request for comment on the episode.

Why did a founder hand an office move to an AI?

Plato had just raised a new fundraising round and was growing the team, so the company needed a bigger office. Farlow called the logistics a headache: insurance requirements, mover coordination and building-manager compliance tied to the new lease.

Farlow already ran several AI agents day to day. Instinct helps him at the gym. Grok Bot takes snack orders at his company and buys them on Amazon. Since he admits he doesn't know much about insurance, delegating the move to Dot seemed like a natural next step.

What did Dot actually do?

Farlow gave it a simple brief: "We're moving to this office. You can read my emails to find out the details. Just get it done." When the agent texted him repeatedly for guidance, he pushed back: "Just please, get it done."

Days passed and the movers still weren't finalized. Farlow checked the agent's activity. His verdict: "It was mayhem."

The specifics of the meltdown:

• Dot sent multiple back-to-back messages to the movers, triple- and quadruple-emailing every tiny update.
• When movers proposed a 9 a.m. slot, Dot replied: "Is that Pacific Time?" It demanded clarification on nearly everything.
• It emailed the insurance company, the current building managers and the new building managers — in some cases all at the same time.
• The new building manager had to start a fresh thread because the original chain grew too large to open.
• Dot shared unnecessary details across parties, telling the movers about Brian, the building manager — information Farlow says they didn't need.

In total, Dot sent around 300 emails to all parties within a few days. The movers eventually caught on that they were dealing with an AI and wrote that they lacked the bandwidth to keep answering its questions. Farlow believes the other parties never figured it out.

What happened after the shutdown?

Farlow said he felt terrible reading the movers' email: "I saw the email from the movers and felt so bad. I'm like, 'What did I do?'" He shut Dot down immediately and says he still owes them a response.

The deal with those movers collapsed anyway — they lacked the necessary insurance. Farlow hired different movers for the actual move, without Dot's help.

He hasn't abandoned AI for the task, but he's changed the workflow. He still uses AI to draft emails, and now reviews everything before sending.

What's the lesson for businesses deploying agents?

Farlow's advice is blunt: unless you're absolutely sure the models will do a good job, you should review everything your agent does. That's a meaningful caveat for companies rushing to hand operational tasks to autonomous tools.

He still sees the episode as a preview, not a cautionary dead end. "The models will get better at this, and I think you're going to see more of this," he said. His closing prediction: soon, it'll be another agent — not a person — on the receiving end of all those emails.