Picture a new intern walking into an organization on their first day. You hand them small tasks, supervised. Once they handle those well, larger ones. And months later, when they have shown again and again that they can be relied upon, real authority. No one hands the intern the keys to the vault on the first day. Authority is not granted all at once; it is earned along a path.
We believe AI agents should travel the same path. An agent’s authority — what it is allowed to do without oversight — should not be a fixed setting at the moment of installation. It should be something that rises over time, with evidence. Authority is not a key you hand over at the start; it is a trajectory the agent walks.
From the bottom of the path, upward
Every path has an entry point. For a new agent, the entry point is the lowest-risk work: tasks where, if they go wrong, the cost to correct is negligible and a human can easily review them. This is the bottom of the path — where the agent shows its work and we measure its behavior.
As it climbs, two things change at once: the risk of the work it is given grows, and the oversight on it shrinks. But this climb is neither automatic nor time-based; it is conditional. An agent moves up a step only when, on the step below, it has repeatedly and stably shown that it can be relied upon. The path runs upward — but every step of it is earned.
Evidence, not promises
This is exactly where this view differs from “trust the AI.” We do not assume trust; we build it. Every step of the path is backed by evaluation: real examples that show whether, at this level of responsibility, the agent’s behavior is stable and inspectable. More authority is the reward for more evidence.
And because the path stands on evidence, it is reversible too. If an agent’s behavior drops at some level — because the data has shifted or the work has grown more complex — we lower its authority. This is not a sign of failure; it is a sign of a healthy system. Just as in a good organization responsibilities are tuned to performance, an agent’s authority should stay live and accountable.
A forward-looking idea
Part of what we describe here is the direction of our design, not the claim of a finished product. The goal is clear: systems that manage authority like a path, not like an on/off switch. This view moves AI from a static tool toward something closer to a colleague that grows over time and earns trust.
Don’t gift authority on day one. Let the agent earn it along the path, the way any professional does. You read a page from the bottom upward; you climb a career step by step; and the authority of a trustworthy agent walks that same upward path.