AI you can let
near real work
The question is never whether AI can do something. It is whether it should do it without asking. NessisOne answers that per seat and per kind of action, and records every answer.
Three answers for
every kind of action
Each seat gets a setting for each of nine kinds of action. Anything you have not granted is not allowed.
The seat cannot take this action, whatever it is asked.
The run pauses and the request waits in your approval queue.
The seat acts on its own, and the action is still logged.
| Action | Research Coordinator | Researcher | Writer |
|---|---|---|---|
| Read company data | Goes ahead | Goes ahead | Goes ahead |
| Read the web | Not allowed | Goes ahead | Not allowed |
| Draft work | Not allowed | Not allowed | Goes ahead |
| Change records | Not allowed | Not allowed | Not allowed |
| Publish or send | Not allowed | Not allowed | Not allowed |
| Spend money | Asks you first | Not allowed | Not allowed |
| Schedule | Not allowed | Not allowed | Not allowed |
| Delegate | Goes ahead | Not allowed | Not allowed |
| Change security | Not allowed | Not allowed | Not allowed |
Built so the rules
can't be talked around
Approve exactly what you saw
When a seat needs permission, the run parks and the request lands in your approval queue. When you approve, that one action runs once, with exactly the arguments you approved, inside the asking seat's limits. An approval can't be reused for something else, and the work picks up where it stopped.
Outside content raises the bar
Once a run has read anything from outside your company, every outward action in that run drops to "asks you first" for the rest of it.
A guard nobody can overrule
The Security Guard on every chart screens outside content for prompt injection and answers to no one in the chain of command. Five layers of defense
Security stays with people
No AI seat can ever be allowed to change security settings. More autonomy is granted by a person, based on a seat's track record, never by the AI itself.
A paper trail for every run
Briefs, work orders, answers, approvals, quarantined content and deliverables are all kept in the engagement ledger.
Enterprise controls
under every department
See the approval queue
with your own workflow
We will walk through a real engagement, pause it on an action that needs your sign-off, and show you everything the ledger recorded along the way.