← sources

"Can act" is not the same as "has enough evidence to act"

source: reddit (reddit:AI_Agents)published 2026-09-26 02:51ingested 2026-09-27 02:35

retained text

Everyone talks about what an agent should be allowed to do

What tools it can use

What data it can access

What actions it can take

But I think there is a question before all of that

Does it actually have enough evidence to act

An agent can have the right permissions and still be working with incomplete or unreliable information

Permission does not make bad information reliable

To me these are different things

Authority means it is allowed to act

Evidence means there is enough information to justify the action

Execution means actually doing it

A system can get the first one right and still make a bad decision because the second one was never checked

So I am wondering where this check should live

Should the agent itself be required to stop when the evidence is not sufficient or should that decision happen outside the runtime

Curious how people building agents are thinking about this

document fetched 2026-09-27 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wqfcy5/can_act_is_not_the_same_as_has_enough_evidence_to