Data Estate Diagnostic
Data is ready or not ready for a specific use case, never in general. A diagnostic that grades the whole estate against a maturity model produces a list nobody funds; one that tests the data each priority use case needs produces a fix sequence you can act on.
- Organizations that lack, or are unsure they have, the right data management practices for AI (Gartner survey of 1,203 data leaders, July 2024)
- 63%
- Organizations that lack, or are unsure they have, the right data management practices for AI (Gartner survey of 1,203 data leaders, July 2024)
- Gartner, 2025(opens in a new tab)
- Global systemically important banks fully compliant with all BCBS 239 risk data principles (Basel Committee assessment, published 2023)
- 2 of 31
- Global systemically important banks fully compliant with all BCBS 239 risk data principles (Basel Committee assessment, published 2023)
- Basel Committee on Banking Supervision, Bank for International Settlements, 2023(opens in a new tab)
- Chief data officers confident their data can support new AI-enabled revenue streams (IBM survey of 1,700 data leaders, 2025)
- 26%
- Chief data officers confident their data can support new AI-enabled revenue streams (IBM survey of 1,700 data leaders, 2025)
- IBM Institute for Business Value, 2025(opens in a new tab)
The problem
In a Gartner survey of 1,203 data management leaders in July 2024, 63% of organizations either did not have or were unsure if they had the right data management practices for AI. Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data1. IBM's 2025 study of 1,700 chief data officers found only 26% were confident their data can support new AI-enabled revenue streams2.
Regulated firms have met this problem before. The financial industry consistently names data challenges as major barriers to further AI adoption, according to the BIS Financial Stability Institute3. The Basel Committee's 2023 review found only two of 31 global systemically important banks fully compliant with its risk data principles, more than ten years after they were published in 2013, and noted that some banks considered themselves fully compliant while still running multi-year data quality remediation4. In January 2026 it said legacy systems, distributed data estates and shifting data lineage still complicate the work5.
Gartner's definition of AI-ready data is tied to the use case: the data must be representative of it, including the patterns, errors and outliers the model will meet1. The EU AI Act sets the same test in law for high-risk systems: data sets must be relevant, sufficiently representative and, to the best extent possible, free of errors and complete in view of the intended purpose6. So a useful diagnostic starts from the use cases you plan to fund and works back to the sources, lineage, quality and access each one depends on.
What you get
- 01
Data requirements per use case
For each priority AI use case, the data it needs written down so it can be tested: sources, fields, history, freshness, labels, and who is allowed to use it for what purpose.
- 02
Fitness tests on real data
Samples profiled from source systems for completeness, accuracy, representativeness and timeliness against those requirements, so findings rest on evidence rather than interviews.
- 03
Lineage and ownership map
Critical data elements traced from source to the point of use, with a named owner for each and the manual steps along the way made visible.
- 04
Governance and control review
Policies, classification, consent, retention and access controls checked against an industry framework such as DCAM or DAMA-DMBOK and the rules that apply to you.
- 05
Ranked gap list and fix sequence
Each gap tied to the use cases it blocks, ordered so the first fixes unblock the most value, with an estimate of the effort and an owner for each.
Where it goes wrong
The failure modes we design against from the first week, and the control for each.
Grading the estate instead of the use case
Scoring every domain against a maturity model gives an average that no one can act on and a backlog too large to fund. The control: tie every finding to a use case it blocks, and leave out what blocks nothing on the plan.
Trusting self-assessment
The Basel Committee found banks rated their own data more favorably than their supervisors did, with some claiming full compliance while struggling with systemic data quality issues4. The control: test samples of real data and trace lineage end to end before writing a finding.
Leaving unstructured data out of scope
Generative AI runs on documents, emails, contracts and call notes, which most data programs never governed. IBM found only 26% of chief data officers are confident their organization can use unstructured data in a way that delivers business value2. The control: include document stores, their permissions, versions and retention rules in the review.
Treating readiness as a one-time project
Data changes after the review ends. Gartner describes AI-ready data as a continuing practice that needs ongoing metadata management and data observability, not a one-off effort1. The control: leave behind automated quality checks and owners for the critical data elements, not just a report.
Standards and rules
BCBS 239, Principles for effective risk data aggregation and risk reporting
Primary text on bis.org(opens in a new tab)Banks are expected to have architecture that supports risk data aggregation in normal times and in stress, with data aggregated on a largely automated basis, and AI data paths should be held to the same standard.
EU AI Act, Article 10 (data and data governance)
Primary text on ai-act-service-desk.ec.europa.eu(opens in a new tab)For high-risk systems, data governance must cover data collection and origin, preparation steps, examination for bias, and the identification of data gaps and how they will be addressed.
DCAM v3 (EDM Association)
Primary text on edmcouncil.org(opens in a new tab)Gives a scored model of 34 data management capabilities and 101 sub-capabilities, useful when you need a benchmark, with assessments that produce auditable evidence for market authorities.
DAMA-DMBOK, 2nd Edition
Primary text on dama.org(opens in a new tab)Provides the common vocabulary for data management functions, so the findings map to roles and practices your data office already uses.
How to measure it
- Share of critical data elements for priority use cases with a named owner and documented lineage
- Quality scores for those elements (completeness, accuracy, timeliness) against agreed thresholds
- Blocking gaps closed per quarter, and the use cases they unblocked
- Time to provision an approved data set to an AI team
- Share of critical data elements with automated quality checks running in production
Questions buyers ask
What is AI-ready data?
Data that is fit for a specific AI use case: it covers the situations the model will meet, including errors and outliers, it is accurate and current enough for the decision, its origin and transformations are traceable, and you have the right to use it for that purpose. The same table can be ready for one use case and unfit for another, which is why readiness has to be tested per use case.
How is a data estate diagnostic different from a data maturity assessment?
A maturity assessment scores your data management capabilities across the organization. A diagnostic starts from the AI use cases you plan to fund and tests whether the data they need is available, fit and governed. It uses a maturity framework as a reference, but its output is a ranked list of gaps tied to use cases, not an overall score.
Do we need to fix all our data before starting AI?
No. Fix the data that your first use cases depend on, and do it in the order that unblocks the most value. Trying to clean the whole estate first is how data programs run for years without a visible result. Each use case you ship should leave behind better quality checks, lineage and ownership for the data it used, so the estate improves as you go.
Sources
Every figure on this page comes from the sources below. We checked each link on .
- 1.Lack of AI-Ready Data Puts AI Projects at Risk(opens in a new tab)Gartner, February 26, 2025 · gartner.com
- 2.IBM Study: Chief Data Officers Redefine Strategies as AI Ambitions Outpace Readiness(opens in a new tab)IBM Institute for Business Value, November 13, 2025 · newsroom.ibm.com
- 3.In data we trust? Emerging policy and supervisory approaches to AI data use in financial services (FSI Insights No 73)(opens in a new tab)Financial Stability Institute, Bank for International Settlements, March 2026 · bis.org
- 4.Progress in adopting the Principles for effective risk data aggregation and risk reporting(opens in a new tab)Basel Committee on Banking Supervision, Bank for International Settlements, November 28, 2023 · bis.org
- 5.Implementation of the Principles for effective risk data aggregation and risk reporting (BCBS 239 Principles), BCBS Newsletter No 36(opens in a new tab)Basel Committee on Banking Supervision, Bank for International Settlements, January 6, 2026 · bis.org
- 6.AI Act (Regulation (EU) 2024/1689), Article 10: Data and data governance(opens in a new tab)European Commission, AI Act Service Desk, 2024 · ai-act-service-desk.ec.europa.eu
- 7.Principles for effective risk data aggregation and risk reporting (BCBS 239)(opens in a new tab)Basel Committee on Banking Supervision, Bank for International Settlements, January 9, 2013 · bis.org
- 8.DCAM: Data Management Capability Assessment Model (DCAM v3)(opens in a new tab)EDM Association (formerly EDM Council) · edmcouncil.org
- 9.DAMA Data Management Body of Knowledge (DAMA-DMBOK), 2nd Edition(opens in a new tab)DAMA International · dama.org