Select the statement that best describes the organization today, one per
chapter. The result shows a level for each chapter and the chapters to
read first. Answers are stored in this browser only.
1. Map Is model use in production mapped?Not yet Staff use models informally; nobody has a list. Foundational The model-assisted tasks are listed with phase, role, and review. AI-ready The map includes data sensitivity and model location, passes the check, and sits in the component register; a row is required before deployment. AI-native The map is reviewed each period and published in the quality documentation. 2. Design How do models assist questionnaire and concept work?Not yet They do not. Foundational A model reviews drafts against the wording checklist; the methodologist reads the flags. AI-ready Question bank matches are proposed with confidence and confirmed; variable descriptions are drafted; acceptance is recorded. AI-native Every questionnaire version is reviewed, acceptance is tracked, and cognitive test findings feed the checklist. 3. Coding How is open text coded?Not yet By coders alone, or by a model with no measurement. Foundational A model proposes codes that coders accept or change; accuracy is measured against the final codes. AI-ready A threshold from the accuracy curve automates a share; a sample is re-coded each round; the rule and results are in the quality report. AI-native Corrections retrain or re-prompt the model; coding runs as a pipeline stage on a local model. 4. Editing How do models support editing and anomaly explanation?Not yet They do not. Foundational Edit failures reach editors with plain-language explanations rated on a sample. AI-ready Proposed values are checked against the rules before editors see them; anomaly explanations cite evidence; decisions are recorded. AI-native Routine and non-routine cases are routed from recorded decisions; verdicts feed the explanation classes. 5. Commentary How is commentary drafted?Not yet By analysts from scratch, or by a model with no check. Foundational The headline paragraph is drafted from the tables and checked by hand. AI-ready The full commentary is drafted with the style guide, every number and claim checked by script, the analyst's approval recorded. AI-native Drafts in every language with the glossary; claims sourced from the series history; acceptance tracked per section. 6. Running Where do the models run?Not yet Wherever the tool is; versions unknown. Foundational A local open-weight model runs inside the production network for one task; the version is pinned. AI-ready Batch jobs with manifests; cost and time estimated and measured; versions recorded per job; the switch-off test passes. AI-native A shared model service with quotas; updates gated by the suites; capacity planned from demand. 7. Assurance How is model use documented and assured?Not yet It is not. Foundational A statement of model use exists for one release with decision points and versions. AI-ready Each task has a suite with required scores; effects on the statistics are measured; the statement is published and checked each release. AI-native Statements are generated from the register and suites; a run is audited end to end each year; report cards are published. 0 of 7 answered Show reading path Reset
Interpretation of the result
Chapter 1 first: the map and its rules are a day of work and decide
where every other chapter applies.
Chapter 3 is where most organizations start in practice; the threshold
rule and the re-coded sample are what make it a production method.
Chapter 7 is the one that keeps the statistics official; a release with
model assistance and no statement is a release with an undocumented
method.