Blog Post for Broadcom Mainframe Software by Eric Newcomer
Mainframes run many significant commercial applications and are a key part of strategic investments for digital transformation, hybrid cloud computing, APIs, and, more recently, a rapidly growing number of generative AI and agentic AI projects.
Organizations place new demands on mainframes every day, such as more quickly processing increasing amounts of low value credit card transactions or delivering cash from ATMs anywhere in the world. Generative AI projects are related because they fundamentally depend on the quality of data available to them.
Modern mainframe systems are more than capable to meet these new AI related demands. New hardware and operating system releases bring speed, flexibility, and power. Database updates improve latency, connectivity, reliability, and resiliency.
API toolkits and gateways that simplify access to mainframe data for mobile and web apps can also be used to access data to train AI models. Hybrid cloud architectures improve cost and performance and add options that improve, simplify, and speed mainframe data access for AI and other purposes.
Modern mainframes also offer high performance CPUs that support distributed virtual machines and native large language model (LLM) training applications.
Mainframe data can therefore be used to train LLMs on the mainframe, in the cloud, on distributed systems, or in any combination thereof, thereby delivering new and significant value to organizations investing in generative AI projects.
Native mainframe API support also enables mainframe applications to participate in key digital transformations such as customer friendly mobile and web applications, partner ecosystems, and monetizing key functions such as banking as a service.
Modernization technologies and techniques reinforce the importance of foundational mainframe systems and data and extend them into key new strategic initiatives, including generative AI and agentic AI productivity enhancements for developers, operators, customer service staff, security teams, and analytic functions.
Why Mainframe Data is So Valuable
Mainframes operate in tightly controlled environments, making them the most secure, reliable and efficient platforms for processing business transactions and persisting the resulting data.
Tightly controlled access to computing resources such as memory, disk, and CPU reduce the chances of transaction failure, and improves the ability to roll back when one does fail, automatically preserving data consistency and integrity.
Processing transactions in distributed and cloud environments create additional challenges because they experience network errors, which makes it more likely a transaction can fail and makes it more difficult to recover when one does fail.
Up to date, consistent transactional data is critical because it reflects the state of the business in real time. Each transactional update represents an increase in revenue, decrease in cost, and can reflect an indication about a product’s success or failure when analyzed for trends.
Collected over days, weeks, or months, transactional data reveals market and business indications that can show the impact of investments in sales, marketing, advertising, product releases, and so on. Transactional data also feeds data lakes and enables essential business support analytics.
Mainframe managed data is also a critical source for a growing number of generative AI productivity enhancement applications, such as customer service, sales and marketing, fraud detection, cybersecurity, and regulatory compliance.
And the more that generative AI applications can consume real-time transaction data the more they can fine tune models initially trained on aggregate data to continually improve personalized interactions with customers and other productivity functions.
Transaction history is one of the most essential items in the customer service toolbox. Getting this information from the mainframe (and other systems) as quickly as possible is essential for smooth business interactions with customers – something that can instill loyalty or cause friction if not done right.
Typically, a business sets goals for sales, profits, and expense control. Changes to the state of the business are measured against these goals. Mining such operational data is how you navigate a business through the ups and downs of a market, such as retail banking or credit cards, for example, where you need good information to make decisions on where and when to invest in a new product (because customers of the existing products are falling off), check whether an advertising campaign is effective, or evaluate regional and international variances.
All these aspects of mainframe data make it one of the most valuable repositories of an organization’s business critical and operational data. In considering strategic business projects such as modernization, mainframe data always plays a key role and must be included seamlessly within the broader IT environment.
The Intellyx Take
Transactions processed and store by traditional mainframe systems is a tremendous source of data, including up to the minute state of a business.
The value of such data increases significantly as organizations invest in generative AI and agentic AI systems for enhancing productivity and service levels. Transactional mainframe data is a veritable gold mine for AI model training and fine tuning.
Modernizing applications for AI abilities is a strategic business imperative across the IT landscape. Modernization for AI requires a foundation of up-to-date hardware and software. Special care and attention to mainframe data is often required because mainframes are unique IT environments.
But mainframe environments have not been standing still. The recent and ongoing evolution of mainframe hardware and software reflects not only technological advancements but also their pivotal role in shaping the digital transformation of businesses.
No aspect of this is more critical to achieving business value than the secure and reliable creation, management, and exploitation of operational data for AI model training. Each organization differentiates itself based on operational characteristics that drive a competitive edge.
For generative and agentic AI projects to successfully enhance a business’s operations, they must be trained on that business’s data, including if not especially mainframe data.
Mainframe software vendors such as Broadcom Mainframe Software invest in their products and services to help you achieve strategic business goals, many of which rely on access and utilization of operational data.
The vendors can help you achieve your goals, but it’s also important for you to understand and manage your investments in IT and keep them up to date and not let them underperform.
You can reach out to Broadcom to schedule a review of your data management best practices, modernization plans, and APIs for LLM training.
Copyright © Intellyx B.V. Broadcom Mainframe Software is an Intellyx customer. No AI was used to write this content. Image by Broadcom Mainframe Software.
