Legacy Software a Problem? AI Can Help.

Most organizations spend 40 to 80 percent of their IT budget just keeping the lights on. That’s not a guess. That’s what the surveys say. When most of your budget goes to maintenance, there’s not much left for anything else.

Falling Behind

You hear “technical debt,” and you tune out. You don’t want to hear another request for more money to fix something that isn’t broken. But outdated systems have real consequences:

• Investors may devalue your company when they see significant technical debt on your balance sheet.
• Legacy systems may not have the capacity to integrate with AI, or anything else that’s new.
• You may not be able to recover from a disaster if replacement hardware or software is no longer available.

Technical debt undermines your company’s value and potential.

Losing the Race

Full rewrites for mid-market systems typically cost mid-six to low-seven figures and take 9 to 24 months. Costs are notoriously hard to estimate. Scope creep, undocumented behavior, and edge cases routinely cause overruns of 30% to 100%. Worse, nearly 4 in 5 modernization efforts fail to achieve their goals.

Full rewrites are where a lot of money goes to die.

And money isn’t the only problem. During a rewrite, you’re maintaining two systems at once. The old one still runs the business. The new one isn’t ready. Your team splits its attention. Feature development stops. The new system starts accumulating its own debt before the old one is even retired.

I had a client who spent 20% of gross revenue for two years trying to rebuild their e-commerce platform. The result? Nothing usable. The new system never worked properly, and the old one kept falling further behind. That was over a decade ago. They’re still on the legacy system.

Getting Lapped

The alternative to a rewrite is fixing things piece by piece. Short cycles, one module at a time, spread your costs and reduce your risk.

The problem? At that pace, your new code is already outdated by the time you finish — if you finish.

Updating legacy code is slow, especially if your team doesn’t have the right skills or the original developers are long gone. Analyzing a 50,000-line codebase can take weeks. Multiply that across dozens of modules and “incremental” starts to feel “glacial.”

In practice, many companies start, then stop—the cost overruns,  business disruption, and slow progress lead to abandonment. You end up falling further behind.

Changing the Pace

Let me be clear about what AI can’t do. It can’t understand your business logic. It can’t decide what to fix first. It can’t replace human review.

AI can make you faster.

I used AI to fix an issue in thousands of lines of code in 30 minutes. A junior developer would have needed a full day and a lot of coffee.

After a client’s chief developer left with zero documentation, the client was stuck. A browser update left their legacy invoicing application unusable. The code was a mess. Maintained by one guy for over a decade with no frameworks, no standards, and no version control. AI summarized and mapped out how the software worked in hours, not days.

By using Anthropic’s Claude Code , GitHub Issues, Microsoft’s Playwright, and a host of command line tools like PHPStan, refactoring was fast and reliable.

AI may struggle to find the right refactoring workflow or toolset to create a truly automated pipeline, but that is where you come in. My loop was not completely until I started using GitHub Issues to track PHP Fatal errors automatically. This allowed Claude Code to see if the changes it just made fixed the bugs.

You probably don’t need to change everything. Most of your code will survive modernization just fine. It’s the small, fragile, tangled, and untested portion that causes all the problems. AI makes that part approachable.

That’s what AI changes. Before, incremental modernization was so slow that upgrading to a newer but still outdated framework felt pointless. Now you can modernize in small, targeted steps. If you have to refactor again in two years, it’s not as big a deal.

Ready to Run

If your team has been talking about technical debt, or if a rewrite proposal is sitting on your desk, start with these questions:

• Are you spending more money maintaining than growing?
• Just how dangerous is your technical debt to your business?
• Can you really afford another false start?
• Do you really need to rewrite everything?
• Have you looked into AI-assisted refactoring to help you get ahead?

Your senior devs may be terrified of an AI agent digging into code they’ve been cultivating like a 100-year-old bonsai. But ask them: are they satisfied just keeping the blood pumping enough to stay alive, or are they ready to set you up to win the race?