For a long time, enterprises have faced numerous challenges in the field of data management, from access control and security protection to data hygiene. Now, as companies race to introduce AI technology, these challenges are further amplified—the effectiveness of AI is highly dependent on a solid data foundation.

"The more things change, the more they stay the same," EY Global Chief Innovation Officer Joe Depa told CIO Dive. "Data has always been a key topic among C-suite executives... and now, this topic is becoming even more prominent."

Despite sustained attention and increased urgency, data gaps still exist. According to a survey released by Gartner in February 2025, nearly two-thirds of organizations lack or are uncertain whether they have the right data management practices for AI.

Experts told CIO Dive that CIOs can help companies avoid wasteful spending and failed pilot projects by identifying key gaps and the best ways to fill them, thereby correcting course in a timely manner. But first, companies should assess existing practices, engage leadership, and plan the path forward.

"The challenge is truly understanding what type of data is needed and being able to make that data available," said Sorin Hilgen, Chief Digital Officer and Domestic CIO at EG America. "No one I know would say, 'Yes, our data is perfectly ready, let's get started.'"

EG America is the fifth-largest convenience store retail network in the U.S. by store count, with brands including Cumberland Farms and Kwik Shop. The company has been working to accelerate its digital transformation, with AI playing a central role.

To alleviate data challenges, the company relies on vendors such as Databricks and Quorso. Hilgen said: "AI has an enormous appetite for data. We are ensuring that accurate data streams from various internal and external systems are provided to the AI platform so it can consume them properly."

The retailer hopes AI can help optimize store shelf and space layouts and predict supply demand by connecting multiple data sets. "We are still in the early stages," Hilgen said, but the system will need access to local events, weather, and demographic data, among others. For example, stores near Little League games would increase water orders to handle traffic, while stores in Jewish communities would increase kosher options. If a blizzard is coming, stores can increase milk supply.

Data hygiene, management, and access are critical for these use cases. Regardless of industry, technology executives are having similar discussions about strengthening data strategies. Aflac Executive Vice President and CIO Shelia Anderson said during a March CIO Dive virtual event that the actual work may be more difficult than initially expected.

"When you start to see data quality—or the lack of data quality—it can make you pull back a bit, so you have to spend some time addressing data quality issues to ensure you can get the results you need," Anderson said.

Common mistakes organizations make

Analysts point out that even with good intentions, technology leaders and their companies often strengthen data strategies in the wrong way.

Gartner research found that organizations typically prioritize three criteria when considering AI-ready data: governance, quality, and performance. "I wouldn't say that's incorrect, but the last two criteria they think about are lineage and data diversity, and both are critical to AI," said Roxane Edjlali, Senior Director Analyst at Gartner. Lack of data diversity can lead to bias in outputs.

"Imagine you are training a resume screening model," Edjlali said. "If you only select resumes submitted by men, then when you receive resumes from women, there will be data bias."

Understanding data lineage is equally critical. "When we look at many of the missteps or wrong steps companies take... they often fail to clearly understand the data models used to train AI," Kristina Podnar, Senior Policy Director at the Data & Trust Alliance, told CIO Dive. "What we cannot risk at this moment is organizations being unable to explain where data comes from."

The Data & Trust Alliance released version 1.0.0 of its Data Provenance Standards in July 2024, grouped by source, lineage, and usage. The standards were developed by 19 member organizations, including Nike, Walmart, American Express, and Pfizer, aiming to establish a unified approach, increase transparency of data sets, and enhance trust and integrity in data and the AI that uses it. These standards were derived from use cases across 15 different industries and were synthesized, refined, and validated by a team of CTOs, data leaders, and other executives.

To promote adoption, the alliance last month partnered with the open source and standards organization OASIS Open to form a technical committee. Five technology vendors, including Cisco, IBM, and Microsoft, co-sponsored the launch of the committee.

"We have brought these standards up to a baseline level... but we need to add water to the mix, stir it, and actually start baking the muffins," Podnar said. The organization stated that standardizing provenance protocols and developing tools for automated verification processes will lead to better management.

"This will set this work apart from historical data governance efforts, because it's not just about what I can tell you... it's about how to do it," Podnar said.

Companies are currently facing widespread misunderstanding of what data readiness means. Gartner said that organizations that fail to recognize the significant difference between AI-ready data requirements and traditional data practices are jeopardizing the success of their AI projects.

"You can't build it once and for all," Edjlali said. "You can't say, 'Oh, we're going to create a data strategy, and once it's implemented and practiced, all data will be AI-ready.' That's not the case, because it's highly dependent on your AI use cases and the AI technologies you use."

Correcting course

No matter how far off an organization's current data practices are, CIOs can help correct direction through a step-by-step approach.

"Even if you just start adopting good practices from now on, then look at historical data, prioritize, and gradually incorporate it, that's a good way to turn the ship in the right direction," Podnar said. "Otherwise, the operation becomes very tedious."

Organizations with clear direction also find it easier to achieve their goals. Edjlali gave an example: a basic AI model needs clean data and few outliers to predict correctly. But if an organization wants to train a model for anomaly detection, technical experts should include outliers, otherwise the model will not be able to identify them.

In the past, organizations worried about whether they had enough data, but now the focus has shifted to having the right type of data.

Companies can also consider using AI to support better data practices, Depa suggested. For example, EY uses synthetic data to experiment with data sets while reducing compliance risks. "We believe this will become increasingly important for building a data foundation in the future," Depa said.