Many IT leaders are under intense pressure to deploy AI quickly; 60% have said their organizations have AI initiatives underway, according to Foundry’s 2025 AI Priorities Study.[1]
Data completeness and quality are critical to success. Enterprises typically have sufficient volumes of data. Much of it resides in the mainframe, which typically houses rich data—such as customer information, transactional data, and supply chain interactions—that organizations need to leverage for strategic AI insights.
However, identifying, sourcing, and accessing that mainframe data has proven difficult. A recent survey conducted by Foundry on behalf of Rocket Software found:
- Only 28% of organizations are fully tapping into mainframe data
- 76% of IT leaders said accessing mainframe data is a major problem
Mainframe data often lives in disconnected silos, which increases the complexity of integrating it with modern applications. This fragmentation also makes it difficult to understand exactly what data the organization possesses.
Complex legacy codebases on the mainframe add to the challenge; extracting and transforming the data becomes cumbersome, especially for organizations that have gaps in mainframe skillsets.
Also, consider that mainframes run critical transactional applications that cannot suffer a moment of downtime without causing serious damage to the business. That means data must be extracted without disruption—and securely given the often-sensitive nature of mainframe data—which presents technical challenges.
In addition, migrating mainframe applications and data to another platform, such as the cloud, is not always a reasonable option. Mainframes run crucial applications, and migration could pose considerable risks for disruption, downtime, and security.
A nondisruptive approach to a modern mainframe
Mainframe modernization should be an ongoing journey, not a single, massive project.
Start by addressing mainframe data blind spots. For example, in most enterprises, the data fabric doesn’t extend all the way to the mainframe. In the past, IT has tried to implement point-to-point replication, exporting bulk data into data lakes and employing extraction, transformation, and loading (ETL) processes that can be expensive and fragile. As a result, organizations have found that mainframe data is stale, and its origin is difficult, if not impossible, to determine. Complex, unstructured data often lies completely out of reach.
Overcome these issues by mapping the data landscape. Automated tools can map data access across disparate platforms to understand relationships between applications and workstreams. This approach enables organizations to rapidly gain a comprehensive view of their trusted data.
Next, use data virtualization to replace costly, inconsistent ETL processes and avoid reliance on large, time-consuming batch jobs. This is a secret sauce for IT to improve performance, because ETL can consume up to 20% of mainframe capacity. A data virtualization solution resides on the mainframe and provides real-time views of data from any application, including AI, eliminating silos and giving organizations an up-to-the-minute view of their enterprise data without moving or even touching the actual data—reducing costs and simplifying the data infrastructure.
Ready for more actionable guidance toward mainframe modernization without disruption? Rocket Software has deep experience here. Its solutions automate the creation of a detailed view across mainframe and cloud data sources, empowering the enterprise to securely and compliantly provide rich mainframe data to AI.
Continue your mainframe modernization journey by clicking here.
[1] Foundry, AI Priorities Study, 2025