CONCEPTUAL RESEARCH · ORDIEN
Automation Capability Index: methodology
How Ordien will evaluate bounded AI capability using reproducible scenarios, independent evidence checks and transparent failure reporting.
Experimental foundation; no live AI benchmark published
The initial local release tests the evaluation harness with synthetic manifests and deterministic probes. It does not establish AI competence at any business Work Type.
Methodology principles
Pin Work, Role, scenario, profile, fixture and grader revisions. Record evaluation dates and complete failure history.
Grade output correctness and evidence sufficiency separately. A completion claim is not evidence.
Report tested conditions, coverage, failures, oversight and measured costs where available. Unknown cost remains unknown.
Controlled tests differ from organizational execution and independently verified outcomes. No universal automation percentage or workforce-replacement claim follows from them.
Publish findings only after methodology and scope review, with supporting runs and visible recency. Sponsorship cannot change grading or suppress failures.
Use permitted fixtures. Private customer Work, Workers, decisions and evidence are excluded from public evaluation intelligence.