How do you modernize old software without stopping the business?
An Access or FileMaker system, home-grown software or a big Excel workbook has been running the business for years, and it's starting to show its age. Here are the signs it's time to act, the 3 ways to modernize it without stopping everything and what AI changes about this work in 2026.
:quality(80))
It's been running for years. Someone built it in Access, in FileMaker or in a big Excel workbook, because no off-the-shelf software did exactly what the business needed. It's been modified often. Today, everything goes through it.
And it did its job. That's actually why it's still around.
But it's starting to show its age. It crashes when 2 people open it at the same time. Only one person still dares to touch it. And the file is now called Orders_FINAL_v3_really_final.
Replacing it is scary, and rightly so: it's the heart of operations. The good news is that there's more than one way to modernize it, and the business doesn't have to stop in the meantime.
How do you know it's time to modernize your system?
It's time when the system costs you more in workarounds than it brings in, or when it depends on a single person. Old software isn't a problem in itself. Software that nobody can evolve is.
The real risk isn't that it crashes one day. It's that everything rests on the one person who knows how it works. Sooner or later, they'll take a vacation, change roles or retire.
A big Excel workbook deserves the same question. Excel is excellent for analyzing numbers. It's less suited to being the central system of a business, edited by 5 people at the same time.
Examples
- Only one person knows how to change it, and nothing is documented
- It only runs on one computer, or on a version nobody has anymore
- Your team can't use it from the job site, the road or home
- It doesn't connect to anything: everything that comes out of it is retyped by hand
- The file crashes or gets corrupted when several people open it
- You hear "the system doesn't allow it" at least once a week
Why can a system that works still be expensive?
Because its cost doesn't show up on any invoice. It's in the workarounds, in the risks you take without knowing it and in everything the business can't do in the meantime.
People in the trade call it technical debt. It's like a line of credit: you moved fast at the start, and the interest piles up. Every change takes longer than the last. Every postponed update makes the next one riskier. And every year that goes by makes the way out a little harder: more data, more workarounds, more things only one person knows.
Then there's what you don't see: the customer portal you didn't build because the system couldn't keep up, the field app that's waiting, the AI you'd like to connect to data stuck in a file. What costs the most is often everything you don't do in the meantime.
Even young software needs upkeep so it doesn't get to that point. We cover it in Without maintenance, your software is at risk.
What are the options for modernizing old software?
There are 3: keep the system and connect it, rebuild it piece by piece or replace it. The right option depends on what's still working in your system, not just on what's wrong.
These options are often combined. You can connect the old system right away to stop the retyping, then rebuild it piece by piece at the pace of the business.
Keep it and connect it
The system does its job well, but it works alone. You link it to your other tools to stop the retyping and, if needed, make it accessible from the field. It's the option that changes the least for your team, but it doesn't solve the dependence on the person who knows it.
Rebuild it piece by piece
You replace one part at a time, starting with the one that hurts the most, while the old system keeps running for the rest. It's often the right choice for a system at the heart of operations that can't stop.
Replace it
Off-the-shelf software already does what you need, or the system is too fragile to touch. You prepare the switchover carefully and change all at once, at the right time. The data, however, has to come along in full.
How do you rebuild a system piece by piece without stopping the business?
By replacing one part at a time while the old system keeps doing the rest. Your team never takes a big leap: it changes how it works one piece at a time.
You often start with the piece that hurts the most, or the one that's easiest to separate: order taking in the field, for example. During the transition, the old and the new share the same data, so nobody has to enter anything twice. The most important thing is the order: you choose each piece based on what it brings and what it risks, not on what's easiest to program.
Every piece replaced is one less piece that depends on the old system. The day nothing is left in it, you turn it off. Often without anyone noticing. That's a good sign.
How does AI speed up taking over old software?
It reads the old system much faster than a person can. That's often the longest part of a modernization: understanding what the old software really does, with its rules that nobody ever wrote down anywhere.
Coding assistants like Claude, ChatGPT, Copilot or Gemini can go through years of macros, formulas and queries to pull out the business rules in plain words. A rule hidden in a 2011 macro becomes a sentence someone on your team can read, approve or correct. They also help transfer the data and compare the old and the new, record by record. And a prototype of the new screens takes a few days instead of a few months, fast enough for your team to try it before going further.
AI reads the old system. A person confirms what it's supposed to do. Because the old system also contains its old mistakes, and exceptions that no longer have a reason to exist.
If Customer.DaysLate > 60 Then Order.Block If Order.Total > 10000 Then Order.Approval = "Name" If Customer.City = "Ville du Lac" Then Order.Discount = 0.05
Why is engineering still necessary, even with AI?
Because AI quickly rewrites what the system does, but it doesn't decide what it should do, and it isn't accountable for your data. Modernizing a system at the heart of operations is still engineering work.
The data is the real project. When we changed Votre Carte Locale's platform, every account, every point and the full history had to come along, without stopping the program. They did. We tell the story of that switchover in Switching platforms without closing up shop, and the full project is on the Votre Carte Locale page.
The rest takes the same rigour as before. AI speeds things up. Engineering keeps the software solid.
All of the data
Every record, every balance, every history has to come along. You check it before the switchover, not after.
Access
Who sees what, who changes what. An old shared file often let everyone see everything.
A plan for switchover day
The right time, what gets tested in advance and how to roll back if something goes wrong.
Upkeep afterwards
New software ages too. Planning its upkeep from the start keeps you from ending up in the same place later.
Frequently asked questions.
How do you replace an Access or FileMaker database?
Start by taking stock of what it really does and who uses it. Then you have 3 options: keep it and connect it to your other tools, rebuild it piece by piece while it keeps running or replace it all at once. In every case, the data has to come along in full.
When should you replace an Excel file with software?
When several people need to edit it at the same time, when important decisions rest on formulas nobody checks or when it holds sensitive data without real protection. Excel is still excellent for analyzing numbers. It's less suited to being the central system of a business.
Can you modernize software without stopping the business?
Yes. By rebuilding the system piece by piece, the old one keeps running while one part at a time is replaced. For a complete replacement, everything is tested in advance and the switchover happens when your customers need you the least.
Can AI rewrite old software on its own?
No, but it speeds up the work a lot. It reads the old code and formulas, pulls out the business rules and helps produce a prototype in a few days instead of a few months. A person confirms what the system should do, and engineering keeps the new software solid: data, security, switchover and upkeep.
How much does it cost to modernize old software?
It depends on what the system does and the option you choose. Connecting generally takes less work than rebuilding, and rebuilding piece by piece lets you spread out the cost. For a simple tool, an app built with the help of AI costs from $5,000.
Your old system did its job. It deserves a well-planned retirement, not a crash on a Friday afternoon.
Start by writing down what it does and who uses it. It's useful whatever option you choose, and it's even more valuable while the person who knows it is still around.
Read next
An idea in mind?
Let’s chat!
:quality(80))
Digital transformationConnect your software or replace it?
Digital transformationSwitching platforms without closing up shop
Digital transformationThe green dividend: a strategic advantage for logistics