An open European standard for reading an organisation's judgment capital while it implements AI: seven pillars, four maturity levels and one floor, at 52.
Four maturity levels, each tab over its own range of the 0–100 score. Open one to read it.
The floor of 52 sits inside Implementing. Certification asks for 75 overall and no pillar below 50.
Like the metre 
set into a Paris wall in 1791, a standard is public. Anyone 
can check against it. This one reads judgment 
.
ISO 42001 covers AI management systems. The NIST AI Risk Management Framework addresses risk identification and mitigation. EU AI Act checklists map obligations. They tell you whether you have the right policies, processes and documents.
Each asks one question about where judgment lives. Each weighs one seventh of the score.
Whether people understand the AI systems they use, and can say what those systems cannot do.
Anchored in Art. 4 and Art. 14(4)(a) of the AI Act.
Whether formal AI rules exist and are enforced: acceptable use, prohibited uses, procurement approval, risk classes.
Anchored in GDPR Articles 5, 22, 25, 32 and 35.
Whether capability is built on purpose, by role: checking output, spotting risk, knowing the ethical limits.
Anchored in Art. 4 proportionality and Art. 14(4)(b), automation bias.
Whether the organisation knows what is deployed, and has human oversight built into it.
Anchored in Art. 4 context of use and Art. 13(3)(d).
Whether the organisation can show it. If the board or a regulator asked for the measures taken, what would it hand over?
Anchored in Art. 4, a duty shown through measures, and Art. 26(2).
Whether someone answers for AI use: accountability, ethical oversight and watch on regulation. The capstone pillar.
Anchored in Art. 17(1)(m), a documented accountability framework.
Where AI may decide and where it may not: written down, watched, and reversible by a person.
Anchored in Art. 14(4)(d–e) and Art. 17(1)(m).
Each level is a band of the score, from 0 to 100.
No formal AI awareness programme. Tools adopted ad hoc by individuals; no usage policy. Staff cannot say which AI tools they use or what the risks are. No measures to show.
Some awareness training delivered. A usage policy drafted but not enforced. A tool inventory started but incomplete. Measures under way, not yet evidenced.
Structured training by role. Policy enforced across the organisation. Evidence of competence documented and auditable. The measures Article 4 asks for are in place along the way.
The floor of 52 sits here.
Continuous improvement, benchmarked against peers, governance built into business processes. Judgment treated as capital: formed on purpose, measured and renewed.
The floor. The point where Developing turns into Implementing. At or above it, the measures can be shown; below it, the gaps can be named. For Article 4, it is the compliance floor.
Certification, first condition: an overall score of 75 or more.
Second condition: no pillar below 50. Both must hold.
Every pillar must score above zero. A pillar at zero is a blind spot in how judgment is formed.
The floor is not a ceiling. At 52 the position is evidenced but fragile; one staff change or one new tool can take it below.
The score is not a pass mark. It is a reading of judgment capital a board can stand behind, and under Article 4 it doubles as evidence of measures.
Many answers become seven pillar scores, and seven pillar scores become one reading. Every step is arithmetic anyone can repeat.
The questions follow the organisation's risk tier: a pharmaceutical company is probed deeper than a design studio. Claims are checked against the evidence held, not taken as declared.
Seven consistency rules test the pillars against each other, so a high policy score with no training behind it is flagged.
Each answer is scored 0 to 100. Answers A and B score 0 to 25 and are flagged as gaps; C and D score 50 to 100.
A pillar's score is the plain average of its questions. Seven pillars, each on its own line of the report.
The overall score weighs every pillar equally, one seventh each. From it follow the level, the floor and the certification test.
Article 4 is now a best-efforts duty under Regulation (EU) 2026/1744, discharged by showing the measures taken. Each of its phrases is evidenced by one or more pillars. Point at a phrase to see which.
The standard sets the governance floor. Above it, Judgment Capital sets a ladder of five rungs, from Blind to Antifragile, for the judgment of the people who check the machine.
The standard is published under Creative Commons Attribution-ShareAlike 4.0. Anyone may use, adapt and redistribute it with attribution. The readiness check and the certification programme are TwinLadder's own.
Read the full standard The licence
The readiness check reads all seven pillars against the floor of 52. It is free once, and included in every plan after. See the plans.
Certification in view? Write to us about certification.
TwinLadder