Roona breaks the work down, searches your documents and the web, asks before committing you to anything, and delivers the finished piece.
Until now, an AI saved you time on one task at a time: you asked, it answered, you moved on. The agent changes the grain. You hand over a mission in a single sentence, put together the tender file, bring me up to speed on a client before Thursday’s meeting, and you go and do something else. Roona breaks the work down, reads what you have allowed it to read, searches the web, cross-checks, and gives you a finished piece of work. At every step that commits you to something, it stops and asks. When you come back, you find more than the result: you find the path, with the name of every document and every link it used.
“Draft our response to the town hall tender.” One sentence is enough. Roona breaks it down on its own: what to read, what to look up, what to write. When the work runs long, it leans on secondary agents, then brings everything back together into a single finished piece.
Sending an email, inviting people to a meeting, deleting a file: none of it happens without you. Roona prepares it, shows you exactly what it is about to do, and waits for your click. You can edit the wording, or say no, and it carries on with the mission another way.
Every mission leaves a readable trail: what Roona read, what it searched for, what it produced, step by step. And every piece of information comes with its origin, the name of the internal document, or the link to the page. You never have to take Roona’s word for it: you open it and check.
While Roona works, every step appears on screen: what it looks up, what it opens, what it writes. You follow the mission as it happens, not just the result at the end. Nothing takes place inside a black box, and you always know where things stand.
A detail to clarify, a direction to correct, a constraint to add: you step in while Roona is working. No need to start again from the beginning. It takes your instruction on board and carries on, and the work already done is kept.
No. Anything that commits you, an email going out, an invitation to other people, a deletion, is shown to you before it happens. You see the recipient, the subject and the exact wording, you change what you want, then you approve. Without that click, nothing leaves. The rest of the work, reading and writing, goes on without interrupting you.
What you open up, and nothing else. Depending on the perimeter you set: your company documents, its memory, your tasks, your meeting notes and your emails, whichever mailbox they sit in, including those provided by internet providers. The perimeter is spelled out when the mission starts, and Roona stays inside it.
Every piece of information is tied to where it came from. For an internal document, you get its name and you open it. For the web, you get the link and you follow it. The mission report also shows the order of the steps: what Roona consulted, what it kept, what it set aside. Nothing rests on Roona’s word alone.
You put Roona back on track. Because the chain of steps is visible, you can see where it took the wrong turn: a document read badly, a lead that led nowhere. You say so, and it picks up from there instead of starting over. And since nothing irreversible happens without your approval, a mistake is only ever text to correct.
The mission unfolding step by step, live: what Roona consults, what it produces, how far it has got. You can write at any moment to clarify or redirect. And whenever an action publishes or shares something, it stops on a confirmation card showing exactly what is about to happen: you approve or you refuse. That holds whatever level of autonomy it has been given.