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.
For a fuller assessment, with indicators and a country pathway, see the
program's AI-readiness assessment framework .
The From assessment to action page maps each
of its dimensions to the chapters and tools here.
1. Find How complete and visible is your catalog metadata?Not yet Many series lack a description, unit, or coverage, or have no page of their own. Foundational Every published series has a page with a permanent URL and a complete record. AI-ready Catalog records are also published as DCAT or schema.org markup, and a catalog API or feed exists. AI-native Search works by meaning as well as by keyword, and metadata gaps are found with AI and fixed under review. 2. Retrieve How can software obtain your data?Not yet Mostly from PDFs, Excel files, or pages that need manual steps. Foundational Open files (CSV or similar) with stable download URLs. AI-ready A documented API with stable identifiers that returns unit, period, and source with the values. AI-native An agent interface such as an MCP server, returning provenance with every response. 3. Understand How is the meaning of each series documented?Not yet Definitions and units are in footnotes or missing for many series. Foundational Each indicator has a definition, unit, method note, and standard geography and date codes. AI-ready Structure and code lists are published in SDMX or DDI, with crosswalks to national classifications. AI-native Concepts are linked to shared vocabularies or an ontology. 4. Ask How do users ask for statistics?Not yet Browsing menus or a keyword search without filters. Foundational Search with filters and example queries; failed searches are reviewed. AI-ready Semantic search with structured results and support for the main user languages. AI-native A conversational interface that answers only from retrieved records, with citations. 5. Trust How are source and freshness shown?Not yet Source and release date are not consistently shown. Foundational Every dataset page shows source, release date, and a suggested citation; a revisions log exists. AI-ready Every API response carries source, identifier, and release date; superseded series point to replacements. AI-native Numbers in generated answers are checked against the record and answers are logged for audit. 6. Evaluate How do you measure whether search and answers work?Not yet There is no fixed test set. Foundational A set of known-item questions is run by hand on a schedule. AI-ready Retrieval measures are automated and reported per language. AI-native Answer-level tests (numeric accuracy, citations, refusals) run before every change. 7. Monitor use How do you track use of your data?Not yet Download counts only. Foundational A recommended citation and identifier per dataset; automated clients separated in logs. AI-ready Publication sources searched for mentions on a schedule, with a table of name variants. AI-native Mentions extracted automatically and reported; AI systems checked for how they cite the organization. 8. Govern How is AI use governed?Not yet No written policy on AI use. Foundational A short policy on permitted and prohibited uses; confidential data kept out of external tools. AI-ready A register of AI components with owners; human review for generated content; read-only agent access. AI-native Inputs and outputs logged; prompt-injection tests; an exit plan per provider. 9. Sustain How maintainable is the system?Not yet Key tasks depend on one person and are not documented. Foundational The production process is documented with backups for each task. AI-ready Open standards, version control, scheduled reviews, and a cost estimate per query. AI-native Cost, latency, and quality are monitored together; small open models are used where they suffice. 0 of 9 answered Show reading path Reset
Interpretation of the result
A Not yet or Foundational result in chapters 1 to 3 comes before
anything else. The later chapters depend on complete metadata, open
formats, and documented meaning.
Chapters 6 (Evaluate) and 8 (Govern) can start early at a small scale: a
question set of 30 items and a one-page policy take a day each and make
every later step measurable and safe.
Chapter 4 (Ask) at the AI-native level, a conversational interface, is
worth starting only when chapters 1, 2, 5, and 6 are at AI-ready.