Gate every AI budget on four data checks

On 26 February 2025, Gartner published a prediction under the name of Roxane Edjlali, a senior director analyst covering data management: through 2026, organizations will abandon 60 percent of AI projects that are not supported by AI-ready data. The same release carries a plainer sentence from the same analyst. If the data has issues, the data is not ready for AI.

Seventeen months later, that prediction is running out its clock inside live budgets, and most of the projects it describes have not been told yet.

The mechanism is quiet. An AI project rarely dies at the demo. The demo runs on a clean extract someone prepared by hand, so the model looks good and the budget renews. The project dies later, in production, when it has to pull from the systems the company actually runs: the CRM that spells one customer three ways, the orders database nobody has reconciled since the migration, the pricing history that lives in a spreadsheet on one analyst's drive. The model performed as built. Nobody had been asked to prove the data was ready before the money moved.

Which puts a specific decision on your desk this quarter: whether any AI project in your portfolio keeps its budget without evidence that its data is ready.

Gartner, press release, 26 February 2025. Scope note below.

Gartner predicts that through 2026, organizations will abandon 60 percent of AI projects unsupported by AI-ready data.

In plain terms: of the AI projects running on data that was never made ready, Gartner expects six in ten to be dropped, not finished. The waste covers the build cost, the team's time, and the quarters spent waiting on a result the data could not produce, on top of the software subscription.

The mechanism Gartner describes is a mismatch. In its account, traditional data management is too slow, too structured, and too rigid for AI teams, data sits in silos across systems, and its uses go undocumented, while AI-ready data must be representative of its specific use case and proven ready through metadata, governance, and quality work. Most organizations have the first kind of practice and assume it covers the second. Gartner's supporting survey, run in July 2024 among data management leaders, found 63 percent of organizations either do not have or are not sure they have the right data management practices for AI. The people who run the data gave that answer about themselves.

Scope, honestly: the 60 percent figure is a prediction, not a measurement, so treat it as direction rather than a decimal. It is scoped to projects lacking AI-ready data, not to all AI projects. The 63 percent figure is self-reported by data management leaders, and Gartner's own materials disagree on the survey's sample size, so this issue does not state one.

The line to use in your next executive meeting, scope attached: Gartner predicts that through 2026, 60 percent of AI projects without AI-ready data will be abandoned. It is a prediction, and it is scoped to exactly the projects nobody has gated.

"Our pilots are producing results. Does that not prove our data is ready?"

It proves something narrower. A pilot proves the model works on the data it was given. In most pilots, that data was a prepared extract: pulled once, cleaned by hand, reconciled by the person who wanted the pilot to succeed. Production is a different test. Production means the model pulls from live systems on its own, every day, without a person in the loop fixing the feed.

The honest limit of this week's evidence is that Gartner's number cannot tell you whether your data is ready. It tells you the base rate for companies that never checked. What settles the question for your portfolio is a per-initiative analysis built around four checks: a written source, a named owner, a current quality check, and a live feed, and that takes ten answers per initiative, not a study.

Complete this before next Thursday. Steps 1 and 2 run inside the AI Data Readiness Analyzer, a free tool on our site:

It runs entirely in your browser, your answers are computed on your device, and nothing you enter is stored or sent anywhere.

Reply to this email with your ready, conditional, and not-ready counts. I read every reply.

Next week: What the fastest teams do with the initiatives that failed the gate and the one thing they refuse to do.

Once a month, the AI Citation Report goes deeper: one long investigation into how AI systems decide which companies get named to buyers.

Elizabeta Kuzevska, Co-Founder, Revenue Experts, AI revenueexperts.ai