IT consulting firm Gartner projects that companies will spend annually more than $850 billion in software in 2023 alone. According to UK-based AI startup Diffblue, up to half that spend will be invested in app modernization (and if you work in an enterprise that means Java). If software is eating the world, I guess you don't want your own market share eaten by the competition.

Two highlights grabbed my attention when I spoke to Diffblue's CEO, Mathew Lodge. 1. Almost every respondent said unit tests were essential for success - people are wising up. 2. Writing unit tests, critical to any modernization initiative, consumes anywhere from a quarter to a half of an expensive developer headcount's time on such a drudgery (but essential task). Mathew, who started his own IT career writing compilers and so feels developers' pains, may be familiar to many VMblog readers. He was a longtime senior executive at VMware in their then-nascent cloud group. His latest startup, Diffblue, uses AI to write unit tests for you automatically, 250X faster than you can. It performs this miracle completely autonomously. Mathew recently walked me through the survey highlights.

VMblog:  What should our readers understand most about app modernization?

Mathew Lodge:  Well a quick read of our 2023 Diffblue Java Application Modernization Survey is a good start. We hired an outside firm to do the research and surveyed 450 organizations in the UK and US. As we expected, Java remains the reigning champion of programming languages that companies rely on to run their business. We now know that app modernization has risen to the attention of the C-Suite. Some highlights:

• Organizations expect to modernize more than 80% of their existing Java applications over the next half decade;
• 97% say unit testing is "extremely" or "very Important";
• 96% consider Java applications to be extremely or very important to their organizations;
• 90% expect to use more automation;
• 87% consider Java modernization a higher priority than other projects.

VMblog:  Why should our readers be interested in your survey findings? 

Lodge:  Since you and I go back many years, we both know that VMblog has long been the must-read for VMware sysadmins, virtualization architects and so on. I know I was an addicted reader. Well, fast forward to today and we see that much of that QA and Ops work has "shifted left" (to use a label that's been around for years now). What it really means in practice is that much more work gets dumped on developers and we're not seeing many organizations commensurately increasing their spend on software engineering headcount. On top of that, developers are no longer seen as cost centers but revenue generators. The heat is on them. As a result, increasingly even the best development teams are finding themselves on spinning hamster wheels having to run faster and faster with ever more work to do. Because guess what else? We don't have enough software developers to hire to help them in the first place - and we never will. That's why Diffblue excited me enough to join as CEO. Our AI is a game changer. Survey respondents (more than 90%) said automation is critical to their app modernization plans. We're here to help with AI.

VMblog:  How does your AI help? Why can't developers just use Copilot or ChatGPT or whatever new AI tools the cloud giants are bringing to market that write code?

Lodge:  No one, and I mean no one - can do what we do already today. Our cutting-edge Reinforcement Learning AI engine is the only solution that completely automates writing unit tests. Today. Right now. We're not a science project or a beta program you can join. We work and we're in production in some of the most demanding enterprise environments like banking, insurance, healthcare, cloud providers - industry verticals where downtime is ferociously expensive or even deadly. The popular LLM-based tools like Copilot and ChatGPT are fantastic in their own ways, but today a human being has to sit down and review every line of code they suggest. Because we all know that ChatGPT, for example, can be "confidently dead wrong." Our customers don't have time for that with unit tests. They need perfection, or as close to it as possible, in the sense that they expect correct, working code every time. ‘It might work' just isn't good enough. So our AI unit tests compile 100% of the time and find regressions. You never need to review and correct the code we create (though we also make the unit tests easily human readable). Best of all, developer teams can now focus on the fun things in coding. It's like getting to stuff your dirty laundry into a washing machine instead of having to hand scrub your clothing.

VMblog:  So what's the elevator pitch for Diffblue?

Lodge:  We're generative AI for code. We make AI work for you, not the reverse. Our autonomous code writing capability has far-reaching potential and will eventually transform how software is written. And in lean times like today, when budgets are being squeezed, our product helps organizations do more with less. In short, our solution called Diffblue Cover already gives enterprises running Java a unique way to leverage the business value of unit testing by allowing them to:

• Autonomously write and maintain unit test suites for entire applications up to 250X faster than humans;
• Automatically refactor Java code to improve testability;
• Slash the time and cost to run tests in CI (Continuous Integration);
• Help development managers understand test coverage and any code risk across their organization;
• Make your developers happier and more productive so you can hire more of the best ones more easily and reduce churn on your own team.

If you're with an enterprise on its own application modernization journey, who wouldn't want this?

##