of surveyed organizations either lacked—or were unsure whether they had—the appropriate data-management practices for AI.
Source: Gartner, 2025Source evidence brief · AI-ready data
Access to data is not the same as readiness for AI.
A Gartner Q&A published on February 26, 2025 warned that organizations are putting AI initiatives at risk by treating conventional data access as sufficient preparation for AI. This Trilivy brief preserves the principal findings, explains their practical meaning and records the original source.
Primary source: Gartner. Lack of AI-Ready Data Puts AI Projects at Risk, a Q&A featuring Roxane Edjlali, Senior Director Analyst at Gartner, published February 26, 2025. View the original publication.
The warning
The data foundation is already stopping AI projects.
data-management leaders participated in the Gartner survey conducted in July 2024.
Survey source: Gartnerof AI projects unsupported by AI-ready data were forecast by Gartner to be abandoned through 2026.
Forecast: Gartner, 2025“Above all, if the data has issues, then the data is not ready for AI.”Roxane Edjlali, Gartner
What AI-ready means
A connector is only the beginning.
Gartner’s argument goes beyond giving a model permission to open a repository. Data must be fit for the intended use, supported by metadata and governed as an ongoing practice. Readiness changes with the use case and must mature as the organization’s AI portfolio grows.
Use-case alignment
The data must represent what the specific AI system needs to recognize, including relevant patterns, exceptions, errors and unexpected conditions.
Qualification and governance
Metadata, sensitivity, interoperability, protection and accountability determine whether data can safely support a particular AI use.
AI-specific data practices
Traditional management practices may need to be extended with vector stores, chunking, sampling, embeddings and retrieval-augmented generation.
Continuous maturity
AI readiness is not a one-time conversion. Metadata, observability, pipelines, assurance and governance require continued improvement.
The five actions
How leaders build an AI-ready data practice.
Align data with the AI use case.
Identify the internal and external sources the specific use actually requires.
Define AI governance requirements.
Determine interoperability, sensitivity, protection and legal or ethical constraints before feeding information into models.
Turn metadata into an active capability.
Discover, enrich and analyze metadata so it can support recommendation, qualification and automation.
Prepare production data pipelines.
Build the datasets and live feeds required by training and operational AI systems.
Assure, monitor and improve the data.
Use testing and observability to detect changes, evaluate quality and refine requirements continuously.
The enterprise implication
AI-ready data requires ownership above individual projects.
As organizations move from experiments to operational AI, data readiness becomes a cross-functional discipline. Executive leadership, data and analytics, IT, security, legal, business owners and AI delivery teams must make coordinated decisions about value, risk, access and accountability.
Trilivy interpretation: this supports the Platform-First doctrine. If every chatbot, agent and application prepares the same corporate knowledge independently, the enterprise multiplies pipelines, metadata, governance work and inconsistency. A shared governed corpus creates one foundation that many authorized AI consumers can use.
This interpretation is Trilivy’s analysis. Gartner did not evaluate or endorse Trilivy.AI, Jumpstart or the Platform-First doctrine.
Source record
Original publication details.
| Publisher | Gartner, Inc. |
|---|---|
| Title | Lack of AI-Ready Data Puts AI Projects at Risk |
| Format | Q&A with Roxane Edjlali, Senior Director Analyst |
| Publication date | February 26, 2025 |
| Survey basis | Gartner survey of 1,203 data-management leaders conducted in July 2024 |
| Original source | Read the Gartner publication |
This page is an independent summary and commentary prepared by Trilivy.AI. It is not a reproduction of the Gartner article. Gartner and its publication titles are the property of their respective owner.