Three strategies for successful mainframe modernization

While mainframes are often dismissed as old technology, for many enterprises they continue to be the cornerstone of a hybrid IT strategy, supporting the organization’s most important applications and business processes. On average, enterprises with mainframes use them to run 56% of their mission-critical applications, according to Kyndryl’s 2024 State of Mainframe Modernization report, which is based on a survey of 500 IT and business leaders. In this research, 89% of respondents said the mainframe was important to their business operations.

If the mainframe is core to your business, then it should not be siloed. And it’s critical that it’s modernized and used to its full potential—especially given the opportunities brought by generative AI.  

Modernization generally takes one of three forms. Our research found that 40% of respondents are modernizing applications while keeping them on the mainframe. Another 35% are modernizing while integrating applications or workloads with the cloud or other platforms within their hybrid IT estate. And 25% say they are moving applications off the mainframe entirely. 

Each of these approaches is equally valid, and each is appropriate in a certain context. There’s no right approach—the goal is to match the right workload to the right platform.

Enterprises reaped impressive one-year ROI from their modernization efforts—between 114% and 225%—depending on the approach taken.

That’s not to underestimate the difficulty of modernization. Senior leaders and even boards of directors often get brought into modernization initiatives, precisely because the applications that run on the mainframe are essential to the day-to-day business operations.  

In our experience with large organizations and in research, we’ve found that three strategies are key in helping enterprises successfully modernize their mainframes and reap the most important benefit of all—sustained competitive advantage.  

Address modernization as part of a complete hybrid IT strategy

Historically, mainframes have been siloed within an organization’s IT environment. That’s no longer necessary or strategic. Case in point: You can do a wide range of modern development on the mainframe. You can use APIs to connect your mainframe to the cloud. You can run containers on your mainframe. The director of an Australian government agency who participated in the survey told us that their team had broken down some of their monolithic applications into microservices, then put them into containers that continue to run on the mainframe. This enables new functionalities and updates without causing downtime or compromising the system’s stability.

Addressing mainframe modernization as part of an overall hybrid IT strategy allows you to optimize your processes and operational insights and make better use of technologies such as AI and automation. A company that wants to check for fraud in its transactions may only be able to test a sampling of those transactions if it collects data at a point-of-sale terminal, sends that data to the cloud, runs an algorithm, and then returns the results. By doing all that on the mainframe, it gains significant speed and can check every transaction.

Technology and business leaders are discovering use cases for AI and generative AI that are well suited for mainframes. Some 86% of those surveyed said they are deploying or planning to deploy generative AI to the mainframe. A Brazilian insurance company that participated in our research is using generative AI to find and analyze complex data relationships within their data sets, helping to inform their underwriting and marketing strategies. A German travel company in our survey uses AI to optimize flight scheduling and crew assignment. By integrating that technology with their mainframe environment, they see the potential to improve operations, the customer experience, and cybersecurity.

The mainframe has the power and speed to train and run large language models. In many cases, the required data is already on the mainframe. For some organizations and use cases, it will make more sense to run the model on the mainframe than to shuttle data to a model running in the cloud.

Take an iterative approach

It’s tempting to try to modernize your mainframe all at once.  

That’s a bad idea. A big-bang approach seldom leads to lasting success from modernization. The original driver for modernization, and the most important use cases, can get overlooked in the scale of a larger effort. And often, a plan to do everything at once actually means that nothing gets done at all, or there is the risk of failing completely. Few want to take the risk.

Instead, let business goals determine the scope of your modernization:    

• What innovation do you want to bring to the business? 
• What capabilities do you want to offer? 
• What technology could support that? 
• Realistically, what options do you have? 
• What might it cost?

You’ll want to consider each application separately, and each workload aligned with that application. 

By working iteratively, you decrease your need to see into the future. If you embark on a three-year modernization program, there’s a good chance the world—and importantly, your business—will look a lot different in three years than it does now. If you take one step, and one business goal at a time, you can adjust as conditions change.

Tackle the skills gap head-on

In our survey, organizations say they’re missing a raft of skills, from AI (43%) to cybersecurity (45%) and mainframe-specific skills (41%).  

Still, enterprises are finding ways to get the work done. They’re upskilling, hiring strategically, and partnering to find the expertise that will help them modernize. Our survey found 77% of respondents are turning to outside partners to fill gaps in their talent pools. They’re investing heavily in cybersecurity and regulatory training, analytics, and AI.  

It’s important to develop both a breadth and depth of skills. At Kyndryl, we’re training 5,000 of our mainframe experts on AI, and we have more than 41,000 hyperscaler certifications. You need to tie multiple platforms and technologies together, from mainframe to cloud and AI.

Generative AI may also be part of the solution. Organizations say they’re short on AI skills, but also see the potential to use generative AI, for example, to translate older languages into newer ones. That will help enterprises better understand their monolithic applications. Enterprises are already using AI to better understand how their data is mapped, and to find dependencies in their applications. An impressive 71% of enterprises in our survey say they are planning or already using insights from generative AI to support mainframe modernization. 

No enterprise, and no technology, remains competitive by standing still. Strategic modernization programs are ensuring that mainframes continue to meet today’s critical business needs and evolve to meet the demands of tomorrow. 

Petra Goude is Global Practice Leader for Core Enterprise & zCloud at Kyndryl.