Enterprises are struggling to unlock data goldmines locked in legacy mainframe applications for business insights, analytics and AI adoption, according to a Rocket Software report. The software company commissioned Foundry to survey 213 data analytics, management, engineering and architecture professionals in May. 

More than three-quarters of respondents said accessing mainframe data and contextual metadata is a challenge and nearly two-thirds have had trouble integrating mainframe and cloud data sources. 

As enterprises build out cloud deployments and ramp up generative AI adoption, executives are eying the rich data deposits locked away in mainframe applications. 

More than half of respondents to the Foundry survey say creating analytical tools and fueling business initiatives hinges on access to mainframe data. But only a little more than one-quarter have the capability to leverage mainframe data assets. 

The mainframe has weathered more than a decade of movement to cloud without losing its grip on the enterprise. Businesses still run roughly 70% of global transactions by value on mainframe systems, according to IBM. 

Rather than migrating the data to the cloud, companies are looking to run compact, task-specific models in hybrid environments, according to Kyndryl research. Bringing the AI on-prem to consume proprietary data rather than piping massive data streams into the cloud helps allay security and compliance concerns, Kyndryl found. 

IT leaders have also turned to generative AI coding tools to help modernize legacy COBOL applications, according to a recent IBM report. However, refactoring applications is generally a better option than full rewrites, which are prone to failure, according to a Forrester report commissioned by Rocket Software. 

Most organizations can tap into core mainframe applications and extract value without a major overhaul — once they’ve identified and classified the data they want.