“Every truth has an asterisk.”
‍
– Roberto Ostinelli, Mechanical Orchard CTO

‍

If you consider the arc of human progress, you might be forgiven for suspending the adage, “If it sounds too good to be true, it probably is.”

To a hunter-gatherer, it sounds too good to be true that most of the world’s population can now turn a handle and get running water.

To a medieval serf, it sounds too good to be true that you can expect most children will reach adulthood. (I mean, the decline in child mortality from as recently as the 1950s is astonishing.)

With my friends Opus and Astra, I suspend the adage on occasion myself. It sounds too good to be true that I can finally wrangle my Inbox and Calendar into submission—yet I have (with a few glitches). It sounds too good to be true that I can create an entirely bespoke financial planning calculator molded to my needs—but it now exists. The limits of what I, as an individual, can do with AI have receded into the far horizon: I am all-powerful, managing my agents.

And so it might be with mainframe modernization. What used to sound too good to be true is now enticingly possible: AI can modernize mainframes in a short enough time to not kill your career and in a way that won’t blow up your operations.

But. (Here’s the asterisk.)

‍

‍
No one actually wants to modernize their mainframe. What people really want is to make changes. Modernization only comes into play when a system stands between them and the change they want to make.

And the ratio of cost to impact has to be as small as possible. Gaining $20,000 more in revenue every month by adding two fields in a form is much more attractive when that change costs $10,000 and takes one week. It doesn’t make sense if the change costs $600,000 and takes eight months, assuming you could get it prioritized within the IT backlog anyway.

Surgical change is the new game. It makes you get to where you want to go faster.

But you can’t move fast when change is hard and risky. Change has to be safe, and safety is where mainframe modernization efforts often stumble. You must be able to trust that the change won’t cause a cascade of unintended consequences, like a collapsing Jenga tower. Our Imogen platform decomposes a monolith into tests that are traceable and trackable, so you know what a change impacts before you make it. Extracted business rules aren’t just summarized and interpreted, they map to the execution of all code paths. If something does go wrong, you can roll back quickly, correct, and move on.

“From years to months” is a fine slogan. For the change you actually need, “weeks” is better — and “safe” is better still.

Best of all, this relieves a persistent tension between the line of business (LoB) and the IT teams: business units want speed and autonomy, while IT wants control and risk management. Assuming priorities are agreed upon (a big assumption), you can get both.

So while asterisks abound, we’re seeing a world in which it’s possible to reconcile historically opposing truths. IT and LoB. Speed and safety. To borrow Emily Dickinson, you dwell in Possibility now. Now go gather paradise.

‍

News & Views

HFS was one of the first research companies to identify the shift to “Services-as-Software,” a delivery model where business outcomes are achieved primarily through intelligent, AI-powered software and automation rather than through traditional, labor-driven human teams or rigid, static applications. They recently published an “HFS Hot Tech” take on Imogen, calling our platform an AI-native example of Services-as-Software, purpose-built — as opposed to the typical entrant, eg. an SI productizing its own delivery book.

Some validation on our “BREs are important but insufficient” from an academic angle: this paper notes that, when it comes to COBOL migration, the constraints are in the data division, so human-readable business rule summaries will leave out lots of them. The authors fix this by teaching a model which constraints to drop; we’d much rather (you guessed it) treat the running system’s behavior as the specification.

AI is now creating parodies of itself. Go on, take a break. Spend five minutes on brainrot with this “music video” on AI and software development.

‍

From the Orchard

Our COO Edward Hieatt will speak at the New York edition of the ISG AI Impact Summit this Thursday, October 29th at 2pm, on how to achieve both speed and safety: by putting verification first. Register here using our referral code, ORCHARDVIP.

HFS estimates that there’s about $18 trillion of trapped value inside the enterprise. At the HFS Autumn Summit in London on November 17, our founder and CEO Rob Mee will participate in this panel, moderated by Melissa Fersht, executive research leader and head of the research innovation team for HFS Research.

It’s not too late to participate in our book giveaway — it ends on October 31st. Don’t forget to choose a title from our bookshelf!

Curious to learn more? Say hello@mechanical-orchard.com.‍

‍

—

View all newsletters here