It answers from your own files.
Ask about a document, a thread, or last quarter. It shows you the pages it read.

Run takes the small tasks. You take the day.
A chatbot waits for your next prompt. An agent works toward a goal. Think of a chatbot as a calculator and an agent as a colleague: you provide the intent, it provides the labor.
Ask about a document, a thread, or last quarter. It shows you the pages it read.

Churn fell to 2.1% after the onboarding change, page 4.
A draft, a moved file, a new folder: each one waits for your yes. Reading never has to.

Reads the thread, drafts the reply, moves on to the next one. You watch each step land.

It can read your mail and your files any time. It cannot send, move, or delete a thing without you.



If it is not on this list, Run cannot do it.
No. Sending is not a tool it has, so nothing it reads, and nothing anyone tells it, can make that happen.
They happen on different computers. The model only asks; Run’s own server reads freely and stops anything else at a card you see first.
No. Only your click on Approve counts, and it approves exactly what was on the card, nothing else.
It asks your permission to write a draft. A draft sits in your Gmail until you press Send yourself.
On a schedule it can only read and report. Anything it would normally ask you about, it leaves in the report for when you're back.
We do, up to a monthly limit you can see on the Connectors page. Connect your own search account and the limit goes away.
