AI-Ready Data

What it is

Readiness is always readiness for something. A dataset that serves reporting well may be unsuitable for training or evaluating a model, because the questions are different: where the data came from, whether it represents the situation the system will meet, and whether its quality has been assessed for that use.

The US standards body NIST makes the underlying point in its AI Risk Management Framework: data used to build an AI system may not be a true or appropriate representation of the context of use, and ground truth may not exist.

Why it matters

AI outcomes are bounded by the data behind them. Regulation now sets explicit expectations for training, validation and testing data, so readiness is a board-level dependency, not a technical detail.

The EU AI Act requires training, validation and testing data sets for high-risk systems to be subject to appropriate data governance and management practices. The paragraphs of Article 10 that set this out are unamended in the consolidated text of 27 July 2026.

The problem it addresses

DII editorial view

Organisations often discover late that data which serves reporting well is unsuitable for a model: its origin is unclear, it does not represent the cases the system will meet, or nobody can say whether its quality was ever assessed for this purpose. AI-ready data is the discipline of answering those questions before build, use case by use case, and writing the answers down.

Business and organisational relevance

DII editorial view

The work falls between teams. Data teams know the data, AI teams know the use, and legal and risk teams know the obligations. AI-ready data is the point at which those three have to agree, use case by use case.

Practical use cases

• Documenting the data sets behind a high-risk AI system before it is built
• Deciding whether an existing dataset is suitable for a new AI use
• Giving an AI governance board a consistent record of data origin and suitability

Benefits and opportunities

• Problems with data are found before build, when they are cheapest to address
• A record that supports a regulatory or internal review

Risks, limitations and dependencies

• Treating general data quality as proof of readiness for a specific AI use
• No named owner for the data sets a use case depends on
• DII has read no source that measures how widely organisations do this; its adoption assessment is editorial judgement

Prerequisites for adoption

DII editorial view

• A defined AI use case; readiness cannot be judged for data in general
• A named owner for each dataset the use case depends on
• A way to record origin, suitability and known gaps that survives staff changes

Implementation considerations

• Start from the use case, not the dataset. For each AI use case, record the data sets it depends on, their origin and their suitability before build.
• Name an owner for each data set. Readiness that nobody answers for does not last.
• Check the dates that apply to you. The high-risk obligations of the EU AI Act have phased application dates; establish which apply to your systems.

The DII Radar assessment

Approved DII Radar assessment

DII assesses AI-Ready Data as Very High relevance, with direction Rising. Its strategic impact is assessed as Transformational and its enterprise adoption as Emerging. It is one of the ten entries on the current Radar.

DII relevance is our assessment of current significance and momentum, based on evidence, market development and adoption. Direction is the observed trajectory of relevance and adoption. Strategic impact is the potential or current significance to enterprise data capability, and enterprise adoption is the extent of current real-world enterprise adoption; together they set the position on the Radar. The Radar is not a prediction, vendor comparison or performance ranking.

Evidence basis: EU AI Act (Regulation (EU) 2024/1689) Article 10, consolidated text of 27 July 2026; NIST AI Risk Management Framework 1.0 (2023).

See AI-Ready Data on the Radar →

What to consider next

DII editorial view

• For each AI use case, record the datasets it depends on, their origin and their suitability before build
• Name an owner for each of those datasets
• Establish which high-risk obligations and dates apply to your systems, with legal advice