The old system nobody wants to open is the biggest risk in your company.
It works. Nobody knows exactly how. And whoever wrote it is no longer with the company. You may have a system like this too. For the first time, AI offers a real way to rebuild it safely.
It's not just old code. It's undocumented know-how
Developers call this legacy code, an inherited system. It still works, but maintaining it gets harder every year. It's often written in an old language, runs on technology nobody supports anymore, and has no tests or documentation.
The biggest problem lies elsewhere. Such a system holds your company's rules that exist nowhere else. How the discount for regular customers is calculated. When an order moves to production. What happens when a customer pays only part. Nobody wrote them down, someone just programmed them once.
With a person
A colleague has worked at the company for 20 years. She knows why some invoices are sent differently and which customer has an exception. She is retiring, and none of it is written down.
With AI
The old system is that same colleague, only in code. It knows everything and explains nothing. And everyone is afraid to change anything, because they don't know what will break three steps further on.
Waiting makes it worse, not better
Three things make old systems a growing risk:
- The people who understand them are leaving. New developers learn other languages and tools.
- Maintenance eats time. Old things get fixed instead of new things getting built.
- Security gets worse every year. A system without updates doesn't meet today's requirements and is an ever easier target for attacks.
Until now, modernization looked like this: a team spent months just reading the old code to understand what it does. Then rewrote it by hand, piece by piece, hoping nothing would break. That's why small and medium-sized companies didn't take it on.
AI shortens the most expensive part: understanding
This is where things change. A language model can go through an old system part by part and explain it in plain language: what each part does, how data flows and where company rules are hidden. According to IBM, this alone shortens the discovery phase from months to weeks.
AI can also:
- translate code from an old language to a new one while keeping the logic,
- propose a plan for rebuilding it in smaller parts,
- add the tests and documentation that never existed.
“A manufacturing company tracks orders in a system written by a contractor 15 years ago. The first step isn't rewriting. AI first goes through all the code and creates a document in plain language: which rules apply to prices, deadlines and order statuses. The production manager reads it and marks what no longer applies. Only then does work on the new system begin, this time with the rules written down.”
Modernization in pieces, not all at once
A successful rebuild doesn't switch off the old system on Friday and switch on the new one on Monday. It happens step by step while the company runs normally. It focuses on three things:
- System structure: instead of one big block, smaller parts that can be changed independently.
- Technology: modern tools that get security updates and can connect to other systems.
- Way of developing: automated tests and small, frequent changes, not big changes once a year.
Where AI's limits are
It's also worth saying what AI can't handle on its own. With tangled logic it can make mistakes. It can produce a translation that looks right but behaves differently from the original system. And it doesn't guarantee the new code will be secure.
So the result doesn't depend only on the model, but on the process around it: tests, human review, verification at every step. AI does the mechanical work faster. Decisions with the highest risk stay with experienced people.
“When we build for clients, we have a rule that solves this problem for the future: code and access are always in the client's name. Our system never becomes a black box that only one supplier can open.”
Ask yourself
Which system would stop working in your company if the only person who understands it left tomorrow?
