On the one hand it’s surprising that 60-year-old mainframes run core portions of so many of the world’s venerable organizations. On the other it’s no mystery: it’s so much easier to patch a system, hack a feature, or do whatever it takes to quickly fix an issue, rather than rethink very old technology to make change easier.

‍

When a system represents an organization’s fundamental business processes, stability is the watchword and we end up being cautious in our approach. Compound these short-term moves over generations, and the result is inevitable: largely stable but creaking systems whose cost of change is sky-high.

‍

The incentive structure for incumbent vendors doesn’t help. Those who have a stable business operating and maintaining mainframes have little motivation to overhaul them as the years go by. The danger of causing problems with a core customer system will always dominate technical decision-making (and rightly so); and maintaining a grip on maintenance contracts will always be the business priority (and understandably so).

‍

For decades these forces have left us stuck in the technological past. Yet, there has been a quietly persistent reverse pressure: an organization’s need for low-cost change. As expectations for high-quality software have increased, organizations need to rapidly deliver new functionality to users. Old technology and fragile codebases mean making change is slow and expensive because of the technical risks involved; the aging out of talent in the ecosystem; and the embarrassment that auditing a system might bring because there is incomplete understanding of its behavior.

‍

Risk vs. Reward: The true cost of delay

Making a business case for modernization is difficult. Yes, the current systems cost tens of millions to run, but these are happily ensconced in the annual budget. Moreover, introducing instability risks far larger costs of downtime, accidental change of system behavior, and potentially unhappy customers. So while—of course!— the organization wants to make change easier, it is hard to quantify the possibility of improved business or customer satisfaction gains against the massive numbers quoted for modernization projects.

‍

Compounding this difficulty is a history of multiple, failed attempts to modernize over the years, wearing down any optimism that the traditional modernization approaches will work the next time. Often it is only government regulation that manages to force attempts at forward progress.

‍

Changing the balance of risk

But over the past few years something has shifted. The reason is simple: we have reached a point where the pressure on the risk equation is too acute to ignore. For more and more organizations there is an outcry for some new way forward, and it so happens that this moment coincides with the surge of hope regarding the promise of AI. Could AI finally mark a breakthrough in the status quo?

‍

At Mechanical Orchard we think the answer is yes…but only if we rethink the overall approach. There’s no shortage of vendors racing to apply AI to mainframe modernization: code translation is the prime target, and using AI to translate code will surely speed it up. But that blithely sidesteps the question of why traditional approaches have been unsuccessful. If our legacy codebases are sprawling, byzantine, and poorly understood, then how can we be certain that a non-deterministic, hallucinating tool like AI will faithfully create the desired results?

‍

We can’t. And so in practice there is a retreat from making use of the latest generative AI. For the sake of certainty and expediency, brute-forcing code translation “with AI” ends up looking more like a literal translation of old code to new, resulting in new code that bears little resemblance to what a human would write and that still is not proven to behave the way it needs to.

‍

A fresh approach to break the stalemate

We believe that an entirely new approach is needed to break out of the old risk equation stalemate and dramatically reduce an organization’s cost and risk of change. The Mechanical Orchard Imogen modernization platform represents a foundational shift that realizes the promise of AI’s potential for modernization by providing a groundbreaking continuous modernization workflow.

‍

It works like this: Imogen records a library of detailed data manipulations performed by the existing system, then uses AI to converge on a new implementation of clean, human-readable code that satisfies those behavioral specifications. Imogen “trusts” the AI by not trusting it at all: instead, we instruct it on what it needs to solve for. With this “verify then trust” model, we can employ the latest aggressive generative AI models. The result is an incremental process in which new components are proved to be equivalent to their legacy counterparts, and are switched out immediately, allowing for immediate low-risk subsequent change.

‍

Imogen represents a new way forward for the industry. And Mechanical Orchard is excited to lead the way to the era of continuous modernization.