Every agent demo has the same moment: the model does something impressive, and somebody in the room asks, “but what if it gets it wrong?” For most teams, that question kills the project.
Our answer is boring on purpose. Agents run on their own until they reach a line you drew. Past the line, a person decides.
The 4% problem
If an agent asks for approval too often, people stop reading the requests. Across two million runs a week, approvals stay meaningful when fewer than 4% of runs need one.
An approval nobody reads is worse than no approval at all — it’s a rubber stamp with your name on it.
Where to draw the line
We start every customer with three rules and tune from there: a dollar threshold, a list of new or unusual recipients, and anything that can’t be undone.
