Earned Autonomy.
Enter one period from your own agent records. You see the share your agents have earned, the count flagged for a person, and the pace your team keeps once people review the work.
How GW Slate™ keeps the run record →The seven deployment controls →
Digital speed, human authority.
When markets moved from open-outcry pits to electronic trading, execution moved to machine speed and the liability stayed with the firm. Safe volume ran at full speed, orders near a limit raised a flag, and firms kept a stop within reach. An agent team reads the same kind of record in its own numbers.
Every order was checked against the firm's limits before it reached the market. On an agent team, the buyer sees which actions ran with no human step and which a person flagged on review.
An order beyond a limit waited for someone with authority. Actions a person flags on review stay with that person.
Every order stayed on record with who sent it. The check shows the count flagged for a person next period.
An accountable person could stop trading at once. A person stays in control of the actions review flags.
Regulators wrote that discipline into rule. SEC Rule 15c3-5 (2010) requires broker-dealers with market access to run automated pre-trade risk controls. CFTC Regulation 1.73 (2012) requires clearing futures commission merchants to set risk-based limits and screen orders against them. Article 14(4)(e) of the EU AI Act writes the kill-switch half into AI law: a ‘stop’ button or similar procedure that brings a high-risk AI system to a halt in a safe state.
Run your own numbers.
Load a month to see how the figures move, then replace the fields with your own agent records for the same period.
Load an illustrative month
The team sees 142 of 168 actions with no human step, 12% of those flagged on review, and 110 actions per hour with review on against 120 with review off.
Agent actions completed with no human step in the period.
Every agent action in the same period.
Share of the actions with no human step that a person later flagged on review.
The same workload with review off.
The same workload with review on.
74.4%
142 ÷ 168 = 0.845, then 0.845 × (1 − 12%) = 0.744
74.4% of all agent actions ran with no human step and held up on review. Read it with the flagged count beside it.
17
142 actions with no human step × 12% flagged on review
Each one was flagged on review. That count sits with a person next period.
91.7%
110 ÷ 120 = 0.917
91.7% of the pace kept once people review the work.
Read the three figures together. Earned autonomy is the share that ran with no human step and held up on review. The flagged count is the work a person takes next period. Pace kept is the share of speed the team still has once people review the work. Results round half-up to three decimals, and every figure stays illustrative until it comes from your own agent records.
Want this measured on your own agent records? Send us a message.
A person can oversee the work, and stop it.
Article 14 of the EU AI Act asks that high-risk AI systems be designed so natural persons can oversee them effectively, including the power to intervene and stop the system. This check shows where that oversight already sits in the buyer's own numbers.
GW Slate™ keeps the run record this check reads, and the record names the person who stepped in. A Manta Graph™ Assessment maps one live workflow before that record is scaled.
The term, defined once.
The buyer reads three figures from one period of their own records: the share of actions that ran with no human step and held up on review, the count flagged for a person next period, and the pace the team keeps once people review the work.
Questions on earned autonomy
01What is earned autonomy?
Earned autonomy is the share of agent actions that ran with no human step and held up on review. Pair it with the pace your team keeps once people review the work, and you can see where agents can take on more and where a person stays in control.
02How do I know when an agent can take on more?
Read earned autonomy beside the count flagged for a person next period. A high share with a low flagged count means those actions held up on review, so that work can move to the agent. A higher flagged count means a person stays with those actions. Pace kept shows the speed the team still has once people review the work.
03Does human review slow the business down?
Pace kept is actions per hour with review on, divided by actions per hour with review off, shown as a percentage. In the illustrative month, 110 actions per hour with review on against 120 with review off keeps 91.7% of the pace. Your own two numbers replace that month.
Move from your own numbers to a measured workflow.
A Manta Graph™ Assessment maps one live workflow on the systems you already run: its decisions, data boundaries and economics, so cost and risk are visible before you scale. Teams already running agents can prove one governed route in 14 days with the Frontier Diagnostic Sprint, then put one workflow under control in four weeks with the Team Control Sprint.