Data preparation is becoming a bottleneck for AI expansion in businesses

A recent study by Iron Mountain reveals that, despite strong interest from organizations in using data to drive results, the lack of a reliable and accessible information base is delaying or even canceling AI initiatives. Over the past twelve months, 86% of surveyed companies experienced delays or disruptions in artificial intelligence projects due to data-related issues.

Time spent on manual data preparation

The research shows that 40% of organizations allocate at least 41% of their AI project timelines exclusively to manual data preparation. This process involves locating scattered files, correcting inconsistencies, and integrating information stored across different environments.

Data fragmentation is identified as a major obstacle. For 56% of participants, data distributed across disconnected systems or departments hinders progress. Another 53% cite outdated formats and physical files as additional barriers.

Only 39% of organizations are considered digitized and structured, with centralized, indexed information ready for use in analytics and AI applications. The contrast is stark: 95% of respondents view using data to improve business outcomes as a priority, yet many still struggle to leverage this information effectively.

AI agents require greater data control

As artificial intelligence agents become more prevalent—capable of performing tasks and interacting with different stages of workflows—the need for accurate and up-to-date data intensifies. To operate reliably, these applications must access correct information, comply with access rules, and allow their actions to be tracked.

However, 70% of organizations cannot effectively track the origin or path of data used by their AI applications. Without this control, it becomes harder to identify errors, investigate results, and mitigate security and compliance risks. Additionally, the lack of clear records undermines the ability to explain an application’s conclusions.

Steps to strengthen data governance

Preparation begins with a detailed mapping of available information. Identifying where data is stored, who is responsible for updating it, and which teams can access it helps establish a more reliable foundation for new projects.

Companies can start by focusing on data directly related to the initiative’s goal, verifying its quality, and defining criteria for tracking results. Governance should be integral to this process: rules for accessing, updating, and using information help reduce inconsistencies and maintain control over applications that increasingly perform tasks autonomously.

While the figures released by Iron Mountain reflect only the organizations surveyed and do not represent a universal estimate, they highlight that maturity in data usage is a critical factor in determining the pace of AI expansion in business.