The COO or CIO of a mid-market IT services firm deciding whether its data is ready for AI should run one AI use case against the firm's own production records and count the records that had to be corrected, instead of reading a readiness survey of its own teams.
TechTarget, a trade publisher, quoted Gartner's senior research director in data management at the Gartner Data and Analytics Summit, in a feature of March 13, 2026: Data could be ready for one AI case but not for another AI case.
Three authors, two at Cork University Business School and one the president of Data Quality Solutions, a data quality consulting firm, published Assessing data quality: A managerial call to action in Business Horizons, an Elsevier journal, in 2020: on average, 47% of recently created data records have at least one critical error, and only 3% of the DQ scores rated acceptable (≥97%). The same three authors' Harvard Business Review article of September 11, 2017 says those figures come from 75 executives who each measured their own department's data quality in the authors' executive education classes in Ireland, not from a survey of organizations, and that no sector, government agency, or department is immune. Its method: assemble 10 to 15 critical data attributes for the last 100 units of work completed by your department, mark the obvious errors in each record, then count the error-free ones.
At an IT services firm the use case can be a service desk assistant that drafts the first reply to a client ticket, which a service desk agent, meaning a person, reviews before it is sent. The assistant reads three records per ticket: the ticket itself, the client's contract record for the service level bought, and the configuration management database, meaning each client's devices and what runs on them. The service desk manager samples the last 100 tickets closed and checks each one: the client matches the contract, the device named exists in the configuration database under that client, and the category and resolution fields are filled. The check runs inside the ticketing system, so no client record leaves the firm before the decision; the cost is a morning of the manager's time, from the service desk budget; the CIO owns the decision.
In the first month the CIO picks the use case, the manager lists the fields it reads and runs the count. In the second the owner of each failed field, the service desk manager for the ticket form, the account manager for the contract record, the engineer for the configuration database, corrects it at the source, and the manager recounts on a fresh sample of 100. In the third the assistant runs on live tickets with an agent reviewing every draft, each client's data terms and the model vendor's contract decide which records may be sent, and the readiness survey, if the firm still wants one, is read next to both counts. An operating partner can ask each portfolio company for the same two numbers on its own use case: the first count of records an error would have changed, and the recount after the field owners fix the source.
Gartner reported on April 16, 2026, in a release Inno-Thought carries word for word, from a survey of 353 data and analytics and AI leaders taken from November through December 2025, that organizations reporting successful AI initiatives invest up to four times more (as a percentage of revenue) in foundational areas, such as data quality, governance, AI-ready people and change management, compared with those experiencing poor outcomes from AI.
