The entry point is a sequence, not a seat
The Permission Ladder
As personal AI capabilities widen what one person can prepare, organizations need better ways to let a newcomer become responsible in public.
From waiting to be assigned a job to learning how responsibility is actually earned
01 / the stalled entrance
Work loses its first rung when no one can touch the real thing.
For much of the application era, an entry-level role was a defined seat. A new worker learned the company’s tools, followed an established sequence, and gradually received more complicated tasks. The route was imperfect, but it gave experience a visible shape. The organization could say what a person was allowed to do next.
That shape is becoming unstable. An AI capability can now produce a credible first draft, summarize a meeting, compare a set of documents, or sketch an operating plan before a junior person has been given a narrow task. This does not eliminate the need for a beginning. It changes what a beginning must accomplish. A person still needs to learn how a real decision is framed, what evidence matters, where a claim can fail, and whose life changes when the work is wrong.
A permission ladder is a visible sequence through which a Stack Person earns increasing authority to observe, prepare, recommend, and finally own a consequential action.
It is not a security hierarchy with nicer language. It is a learning design. Each rung puts a person in contact with a real outcome while keeping the level of authority plain to everyone involved. The point is neither to protect senior people from help nor to hand an automated result a title it has not earned. The point is to make responsibility learnable again.
02 / four rungs
Permission should expand with contact.
These are not job titles. They are different relationships to an outcome, each with a legible reviewer and a chance to see what the work changes.
- 01 / OBSERVE
See the actual decision.
Let the new person see the brief, the available evidence, the disagreement, and the result after it reaches the world. A stack can prepare questions or annotate the material, but observation matters because it reveals the consequence that no generic tutorial contains.
- 02 / PREPARE
Build a contestable first pass.
Give the person a bounded piece of the real work: organize evidence, draft alternatives, identify missing conditions, or make the handoff clear. Their stack may accelerate the first pass. The reviewer’s job is to show which parts are useful, weak, incomplete, or unsafe.
- 03 / RECOMMEND
Make a claim someone can challenge.
Ask the person to state what they think should happen and why. They should name the tradeoff, the uncertainty, and the person who has authority to decide. This rung converts polished output into accountable judgment without pretending that recommendation is the same as commitment.
- 04 / OWN
Carry a bounded result.
Only now does the person make or lead a consequential action within a clear scope: send the proposal, run the meeting, release the change, or own the client follow-up. A reviewer remains available for escalation, but the person can see their authority and the consequences joined together.
03 / a manager’s practical change
Do not mistake faster preparation for a finished apprenticeship.
Imagine a small operations team that has received a messy request from a customer. In the old arrangement, a manager might give the whole case to the most experienced employee because explaining it would take too long. In a bad new arrangement, someone might paste the request into an AI system, receive fluent text, and send it before anyone has understood the real condition.
A permission ladder creates a third route. A newer colleague observes the manager’s first conversation and sees which facts are material. They prepare a short case record with their stack, separating known facts from assumptions. They recommend a response and name the question that still needs a manager’s judgment. On the next similar case, they may own the response within a stated boundary.
The organization has not slowed itself down for ceremony. It has turned repeated work into a way of producing more capable people. The AI capability has a proper role: it helps make preparation cheap enough that more people can get near the real work. It cannot be the person who learns from the consequence or answers for the decision.