I Let OpenAI's Dots Run My Inbox for a Week. It Missed the Point.
The autonomous agent cleared my inbox. It also cleared the emails I actually needed, and learned nothing about why.
I gave Dots, OpenAI's always-on agent, access to my inbox with a simple goal: keep it at zero. For three days it was magic. Newsletters sorted, receipts filed, the obvious spam gone. Then it archived a client's question because the subject line looked like a newsletter, and I only noticed when the follow-up arrived.
That's the problem with autonomous inbox agents. They're great at the volume and blind to the weight. A human skims and knows the weirdly-worded email is the important one. Dots saw a pattern and acted on it, and the pattern was wrong for that one message.
The second issue is memory of the wrong kind. Dots remembered my folders and my filters. It didn't remember that I always answer Sarah personally, or that invoices from one vendor are urgent. Those are the judgment calls that make inbox zero useful instead of just empty.
I kept it, but I changed the rule from 'clear it' to 'draft the reply, ask me.' The autonomy that sounded great is the part I trust least. The lesson: an agent that does the boring 90% is a gift; an agent that decides the important 10% without you is a liability wearing a productivity badge.