Defined term · v1
Autonomy Ladder (L1 → L4)
Definition
The Autonomy Ladder is a four-level maturity model — L1 drafting, L2 gated shipping, L3 customer-facing, L4 full autonomy — for grading how much of a company's operation runs on agents, with the level fixed by what is really done, not by intent.
As of July 2026
The ladder reads like an instrument, not a promise. The level is fixed by what is really shipping — the pointer stays at L1 until the next rung is genuinely done, and the customer-facing rung stays labelled in rollout until it is connected to live channels.
We are not there yet
- L1
Drafting
LiveYou are hereNamed agents draft the routine — pricing reviews, content, competitor scans, weekly plans. Every draft waits for a person.
- L2
Gated shipping
In rolloutNothing reaches a live site without passing the full automated test gate first. Deny-by-default, tested — being widened company by company.
- L3
Customer-facing
In rolloutThe engine that would talk to customers exists in code, but it isn't connected to live channels. Until it is, it says in rollout, not live.
- L4
Full autonomy
TargetTarget — fully autonomousEvery company running end-to-end on agents, with a named human on the one gate that must stay human: the irreversible decision.
Changelog
- v1
July 2026
Initial four-level model — L1 drafting, L2 gated shipping, L3 customer-facing, L4 full autonomy. The pointer sits at L1 and only moves when the next rung is really done.