Tech debt, AI, and the future of software engineering: Lessons from GartnerAPPS 2025

Bryan Friedman
|
June 18, 2025
Real talk at GartnerAPPS
Contents

Key Takeaways

Not long ago, “application innovation” mostly meant choosing between cloud providers or the latest dev framework. But this year at the Gartner Application Innovation & Business Solutions Summit, the conversation had shifted. While there was plenty of talk about what’s coming next with generative AI, intelligent agents, and emerging architectures, there was also a clear focus on how to manage, modernize, and maintain what you’ve already built.

Across breakout sessions, analyst keynotes, and expo hall conversations, a few key themes kept popping up: AI-powered development, technical debt, and modernization. I’ll share some of my key takeaways in this blog.

You can also catch Code Remix Weekly where I break down the event and share some insightful moments from my time there. 

Tech debt gets real

Technical debt has always loomed large in the enterprise, but this year it felt like engineering leaders were finally ready to move from acknowledgment to action. According to Gartner, 25% of engineering time and budget goes toward managing tech debt, yet fewer than half of organizations feel they’re doing it effectively.

What stood out to me wasn’t just this data, but the urgency I heard from folks. More and more engineering leaders are starting to realize that they can’t scale innovation while operating on top of outdated code and fragile systems.

Gartner’s guidance focused heavily on measuring and prioritizing tech debt. They recommend evaluating it across three dimensions:

  • Risk – What’s the potential threat if this isn’t fixed?
  • Impact – How many systems or users are affected?
  • Cost – How expensive is it to remediate, and what’s the opportunity cost of waiting?

To help with this, they reference their PAID assessment (Plan, Address, Ignore, Delay) as a framework for deciding what to tackle first. (What’s a Gartner toolkit without a four-quadrant matrix?) 

Most importantly, they advocate for building a culture of continuous remediation, not just periodic cleanup. That idea came up often in booth conversations too. “Eliminate tech debt” turned out to be one of the most effective phrases we could’ve used on our sign. It pulled people in who were looking for actionable, scalable solutions to this specific problem.

While there was no shortage of advice on identifying, measuring, and prioritizing tech debt, fewer sessions and solutions offered concrete ways to actually resolve it, especially at scale. That’s where Moderne resonated most. We were able to show how our platform automates large-scale remediation, using codified rules to make real progress on legacy systems.

App modernization vs. cloud migration

Modernization was another hot topic that Gartner defines as “building for tomorrow.” Unfortunately, many organizations still mistake cloud migration for modernization. Simply lifting and shifting legacy systems to the cloud without refactoring doesn’t solve much. As Annie Hodgkins put it during a keynote, “That’s not modernization—that’s just relocation.” Similarly, she explained that simply bolting generative AI on top of an old existing stack doesn’t magically make an application intelligent or future-ready.

In one standout session, Gartner’s Matt Brasier offered a practical path forward. His modernization playbook emphasized:

  • Rule-based code transformation for consistent, accurate refactoring
  • Avoiding reliance on AI alone due to its nondeterministic, lower accuracy outcomes
  • Hybrid AI tooling where LLMs assist with identifying targets, but structured rules drive the actual changes

It’s a smart strategy: AI-enhanced, but not AI-dependent. This aligns closely with what Moderne offers—a structured, automated approach to code remediation and modernization.

AI assistants and agents: Inflated expectations

AI was everywhere at the conference. Nearly every vendor on the expo floor claimed to have an AI-powered solution, whether for code generation, testing, documentation, or full-blown “intelligent agents.” But behind the hype, there were plenty of efforts, especially from analysts, to set expectations and ground the conversation.

Joachim Herschmann warned of “AI-agent washing” where vendors market automation as intelligence without delivering on the promise. That theme came up in several sessions. In a presentation on the maturity of AI-related technologies, Gartner’s Arun Chandrasekaran reminded us that most of these tools are still early on the Gartner Hype Cycle, meaning we must be wary of Amara’s Law:

“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

That doesn’t mean AI can’t provide value today. There are real, emerging use cases where AI assistants and agentic tools can offer meaningful support:

  • Documenting legacy systems
  • Extracting embedded business rules
  • Validating changes and running regression tests

But the key, according to Herschmann, is keeping developers in the loop:

“AI agents shouldn’t replace developers. But developers who know how to work with AI agents will replace those who don’t.”

That evolution of the developer role was a recurring theme. As AI takes on more of the boilerplate, developers are being asked to step back from syntax and step up to orchestration. Gartner’s Frank O’Connor put it best:

“The era of developers as code producers is ending. In the age of AI, developers are builders, assembling the right solutions while AI handles the implementation details.”

This shift doesn’t mean letting go of fundamentals. In fact, several speakers cautioned against overreliance on AI coding assistants without strengthening core engineering practices like understanding abstractions, architecture, and testing. AI can speed things up, but only if teams remain accountable for what gets built and how it’s maintained.

Conversations that stuck with me: Change is hard

One of my favorite parts of the summit was talking to folks on the expo floor and hearing real-world challenges from engineering leaders across industries. Here are some examples of what we heard again and again at the booth:

  • “We’re still on Java 7.” Hundreds of developers, millions of lines of code, still stuck on outdated versions and unsure how to safely upgrade.
  • “We spent 1,000 hours updating Spring code, only to fall behind again.” Manual migration efforts just can’t keep up with the pace of change.
  • “We have COBOL systems no one fully understands.” Teams aren’t necessarily looking to fully rewrite, but they need help understanding what’s worth keeping.

Whether it was Java, .NET, JavaScript, or COBOL, the throughline was clear: teams are ready for smarter, safer, and faster ways to modernize. They’re increasingly looking for solutions like OpenRewrite and Moderne that can automate meaningful parts of that process.

Final thoughts: Modernizing at scale is a must

If this year’s Gartner Summit made anything clear, it’s this: the old ways aren’t scaling anymore. Leaders are done hand-waving tech debt. They’re ready to fix it. Developers are done fighting their tools. They’re ready to evolve. And AI isn’t a nice-to-have. It’s quickly becoming part of the core engineering toolkit.

For more insights and stories from the summit, check out the replay of Code Remix Weekly here. And if you’re ready to modernize, Try Moderne.