๐Ÿค–Humanoid Autonomy Compass

How closely does your autonomous system need to be supervised?

A humanoid robot on a shared floor needs different controls than an agent drafting copy. This compass derives the fitting oversight regime from three properties you can actually check, then turns that verdict into a concrete, standards-referenced compliance checklist you can work through and export. Every step of the derivation is shown, and the model is grounded in publicly available standards.

Run timeโฑ๏ธHow long does it act before a human sees the result?
Environment๐ŸŒHow predictable is the context it decides in?
ReachโšกWhat can break in the worst case?

1 Describe the deployment

Pick an example as a starting point, or set the values yourself.

secondsminuteshoursa shiftdays
stableshiftingvolatileunpredictable

2 Where the deployment lands

A continuous risk field. Vertically the autonomy window A, horizontally the oversight regime K. The surface shades from acceptable to not defensible; the marker sits at your deployment with its recommended regime. Hover to read any point.

Autonomy window A  ยท  longer ↑
Oversight regime K  ยท  tight ←  → loose
acceptable watch elevated critical not defensible

3 Recommendation

Derived from run time, environment, and reach.

Autonomy window
A2
Recommended oversight regime
K3

4 Compliance checklist

The controls below are the ones this deployment triggers, mapped to their source clause in each framework. Set a status per control and add evidence; the readiness score and gap count update live. Only controls relevant to your profile are shown.

0%
Indicative mapping, not a certification or legal advice. Clause references point to the public structure of each framework and help you locate the relevant obligation; the authoritative wording is the standard itself. Confirm scope (for the EU AI Act, whether your system is high-risk) with a qualified assessor.
Classes, calculation, and normative sources
ClassAutonomy window A: unsupervised run time in one stretch
A1Single step, every result is reviewed before it takes effect
A2Short action chains in the range of minutes
A3Multi step tasks running up to roughly an hour
A4A full shift with no intermediate check
A5Continuous operation across days
ClassOversight regime K: from tight control to loose review
K1Every action approved in advance by a named person
K2Approval required for critical and irreversible actions
K3Continuous monitoring with the ability to intervene at any time
K4Oversight by exception, alerts only on deviation
K5After the fact review based on logs
StepCalculation
Autonomy window Ataken directly from the run time slider, A1 through A5
Risk load R(environment level minus 1) plus reach points: read only 0, changes state 1, moves physically 2, can injure people 3
ToleranceR 0 to 1: residual risk up to "elevated" is acceptable. R 2 to 3: up to "watch". R of 4 or more: only "acceptable".
Recommended Kthe loosest regime whose risk cell still stays within tolerance
Edge caseif even K1 exceeds tolerance, the deployment is not defensible as configured. Shorten the run time or narrow the reach.
Normative sourceHow it is used here
NIST ALFUS frameworkautonomy as the interplay of task complexity, environmental complexity, and human independence
NIST AI Risk Management Framework 1.0Govern, Map, Measure, Manage subcategories as checklist items
EU AI Act (Reg. 2024/1689), Ch. III Sec. 2high-risk obligations, Art. 9 to 15, with Art. 14 human oversight
ISO/IEC 42001:2023AI management system clauses and Annex A controls
ISO 12100 / ISO 10218 / ISO/TS 15066 / ISO 3691-4machinery and robot safety for systems that move or can injure
SAE J3016analogy for graded autonomy and handover to a human

The weightings and the control mapping are a reasoned engineering choice, not a normative text. Both live in the source and can be adjusted without changing the structure.