Why “big bang” transformations will never work and what to do instead
It’s tempting to think of the mainframe as a relic – an aging system that has no place in a modern enterprise. But just like the microwave is still a fixture in high-end kitchens, the mainframe still has its place: high-volume, low-latency transactions. Where speed and consistency matter more than flexibility or decisioning.
Think:
- ATM transactions
- High-volume trading
- Batch record updates
But here's the trouble: Mainframes handling workloads don't always fall into those categories. Enterprises, at one point, had to use what was available to run high-touch, complexity-laden applications. Mainframes weren't ideal, but they were available.
Now those same organizations are moving processes to the cloud, and looking for ways to embed automation, accelerate with AI, and get more from their data. And they are finding themselves trapped by the mainframe – legacy code powering opaque processes, black boxes from decisions and policies that are missing explicit documentation, data in a closed loop.
Over 75% of modernization projects fail.1 Why?
The solutions to mainframe dependency have come in a few different forms.
The technique: Replatform: Rebuild the entire mainframe estate, painstakingly, line by line, in the cloud.
The problem: It is an incredible amount of work, reliant upon teams of consultants that are prone to human error. And the result is just the same outdated builds replicated in a new place.
The technique: Wrap and renew: Build an agile, adaptive UI layer on top of mainframe applications to enable some connection to modern systems and improve usability.
The problem: Old issues remain below the surface, and maintenance, upgrades, and process changes remain a massive lift. You aren’t solving the problem, only obscuring it.
The technique: Lift-and-shift: Relocate mainframe workloads to the cloud environment with minimal modification to the existing codebase.
The problem: It does little to modernize the application itself. The same monolithic structures, embedded business rules, and operational constraints follow the system into the cloud.
These and other traditional methods don’t get at the reality of mainframes: That some pieces are fit to stay where they are, while others should be extracted, rebuilt for the cloud, and retired from the mainframe. But the nature of legacy mainframes makes that incredibly difficult to pull off. Business logic, data access, and workflows are deeply embedded in the codebase; boundaries are blurred; and data, logic, and UI are tightly coupled.
Modularity is the key to true legacy transformation
Focusing on cloud-friendly business applications – things like claims processing, payment exceptions, account management, fraud handling, change management, and much more – while leaving core transactional processes alone is a recipe for success. How is that possible? Through AI.
AI-powered code analysis can map and isolate the business logic buried inside COBOL programs. But here’s where things get interesting: While some code analysis tools aim to translate COBOL into Java or HTML, the true modernization step is to transform COBOL into documentation that can be utilized for a net-new, reimagined application built for purpose in the cloud.
Once you accelerate discovery and analysis from months to days – or even make such discovery possible in the first place – modular application modernization goes from being another undefined, far-fetched project to being a value-add, iterative, repeatable, and scalable process. Each modernized mainframe application can go live in a span of weeks and start delivering value in a product environment right away while the team identifies the next target.
A factory for mainframe modernization
In this way, modernization projects and the IT stakeholders, vendors, system integrators, business users, and executives who run them become miniature factories unto themselves – pumping out cloud-native, SaaS applications one after another. After the first few successful modernizations, the process becomes repeatable. Teams build a pipeline: analyze, reimagine, deploy, repeat.
Modernization shifts from a once-in-a-decade transformation into an industrialized process.
Analyze the code tied to a particular application, reimagine that as a cloud-ready, AI- and automation-powered workflow, and deploy to a future-proof architecture, connecting to or migrating systems of record and integrations as needed, all in the span of a few months. Rinse and repeat.
Through hyperscalers like AWS, mature transformation practices like Accenture or Infosys, and with the help of the AI-powered workflow design workbench Pega Blueprint™, enterprise CIOs can approach the mainframe modernization projects they’ve held off on with minimal risk, massive upside, and rapid iteration.
1 Source: Survey: 79% of Application Modernization Projects Fail