Every sensitive action is checked against who is asking before it runs. When trust is established, it acts. When it isn't, a person decides. Every tenant is isolated and every action is logged.
Before a sensitive action runs, DaemonLayer checks who is asking. If none of the first three checks says yes, a person approves it, and that approval builds trust for next time.
"Give Jane full access to Tom's mailbox."
"Please add Tom to the Finance group."
"Reset my password."
"Give me access to the Finance mailbox."
Each workflow carries a risk level. A known approver can only authorise it if their score clears that level's threshold.
Some rules can't be switched off by anyone, including us. They are why the rest of the automation can be trusted.



The audit trail is append-only and kept per tenant. It records actor, action, resource and outcome, and Microsoft 365 changes are stored with their before and after values.
In the EU. AI processing runs on models in EU data zones.
No, neither by DaemonLayer nor by the model providers.
The target's manager, a known approver above the risk threshold, or the user themselves for eligible workflows. Anyone else needs a person to approve.
It waits for human approval with full context. Nothing runs.
Each approved request adds 15 to that requester's score, up to 30 in total. Discovered approvers are refreshed weekly.
No. Credentials are encrypted on entry and never exposed again.
Yes. Every action is in an append-only audit log, including before and after values for Microsoft 365 changes.
Start read-only, watch every decision in the audit log, and switch on actions when you trust them.
No credit card. No sales deck. Cancel anytime.