Modernising business software: what AI really changes for an SME
Back to the blog
Artificial Intelligence

Modernising business software: what AI really changes for an SME

Jérémy
Jérémy
Founder & CTO
31 October 20267 min read

Cisco wants to rewrite its old code with AI. What this changes for an SME whose business relies on ageing business software, and how to modernise it without breaking everything.

Modernising business software means putting it back on a maintained technical foundation without losing what it does well: the business rules, the data and the teams' habits. AI significantly speeds up the most thankless part of the work, reading and rewriting old code, but it does not decide for you what to keep. For an SME, the right method remains gradual: an inventory, tests on the existing system, then a module-by-module migration.

The subject took on a new dimension this week. Jeetu Patel, President and Chief Product Officer of Cisco, told Fortune magazine that he had given unlimited AI tokens to Paddy Krishnamurthy, a 27-year-old engineer who joined Cisco when it acquired his start-up EzDubs in 2025. His mission: overhaul the group's legacy code. The stated goal is to process 98% of this code in 24 months, with a team of two people and four weeks to prove that the method holds up.

What does Cisco's announcement really say?

It is a goal, not a result. No figures have been published on the volume of code involved, on the tests planned or on what "processing" 98% of the code means. The team must first show, within a few weeks, that the approach works on a real scope.

For an SME, the figure matters less than the signal. A group of this size now considers that rewriting old code with the help of AI is feasible at scale. The same tools are available to a twenty-person company whose business runs on an application developed ten or fifteen years ago.

How do you know whether your business software needs modernising?

A few situations come up often among the managers who talk to us about their internal tool:

  • The language or framework is no longer maintained, for example a PHP version missing from the list of versions still supported by PHP, and therefore without security fixes.
  • Only one person still knows how to make changes to the tool, and they are not always available.
  • Every small change breaks something else, for lack of automated tests.
  • The tool does not work on a phone or does not communicate with your other software, such as accounting or e-invoicing.
  • The server hosting it has not been updated for years, because nobody dares touch it.

One of these points on its own does not necessarily justify an overhaul. When two or three add up, the risk to the business becomes real. We have detailed the case of outdated PHP versions in our article on the PHP2Shell vulnerability and the PHP update.

Old abandoned control room, with two consoles covered in dials, empty chairs and a wall of clocks

Photo: Alexander Gluschenko on Unsplash

What can AI do in a modernisation, and where does it stop?

What it does well

Today's coding assistants read a file of several thousand lines in a few seconds and summarise its role. They find the places where a calculation rule is applied, suggest a translation from one language to another and write tests that describe the program's current behaviour.

This last point is the most valuable. Before changing anything, you capture what the old software does, so you can then check that the new one gives exactly the same results. A developer who used to spend days understanding a poorly documented module now gets a first overview much faster.

What it does not replace

AI does not know why a particular discount only applies to certain customers, or whether that exception is still intended or dates from a customer who left long ago. Those answers are in your teams' heads, not in the code.

It is also wrong with great confidence. Rewritten code that compiles and survives a demo can hide a calculation error or a vulnerability. This is the limit we described in relation to vibe coding for an SME: producing code has become easy, reviewing it and taking responsibility for it is still a profession.

Then there is the question of data. Taking over ten years of customer, order or invoice history requires checks that nobody can skip, with or without AI.

How do you modernise business software step by step?

1. Take stock of what exists

We list the screens, the automated processes, the imports and exports, the users and what each of them really does with the tool. Often, some of the functions are no longer used by anyone. Those do not need to be carried over.

2. Write tests on the old software

This is the safety net for the whole project. For an invoice, a quote or a schedule, we record what the old tool produces from real data. AI helps a great deal in writing these tests, provided a developer checks that they cover the cases that matter.

3. Choose between a complete rewrite and a gradual migration

Rewriting everything in one go looks simpler on paper. In practice, the company is left for months with a frozen tool, then switches over all at once with all the risks that entails. A gradual migration replaces one module after another while the old system keeps running. For an SME that cannot stop, this is generally the safest route.

4. Migrate the data with checks

Before each switchover, we compare the old and new databases on simple totals: revenue per month, number of active customers, outstanding balances. If a discrepancy appears, we understand it before going any further.

5. Document and train

A successful modernisation leaves behind readable documentation: how the tool works, where the data is, who can make changes to it. That is what avoids ending up in the same dead end ten years later.

Should you modernise the existing tool or switch to an off-the-shelf product?

The question comes up in every project. If your software automates know-how specific to your company, replacing it with a standard subscription often means bending your ways of working to fit the tool. If it reproduces a commonplace function, such as payroll, an off-the-shelf tool will cost less to maintain. We compared the two options in our article business software or SaaS.

When the tool is worth keeping, modernisation results in refurbished custom software: the same business logic, a maintained technology, up-to-date hosting and access from a browser or a phone.

How much does modernising business software cost?

There is no fixed price, because everything depends on the size of the code, the state of the data and the number of connections with other software. AI mainly reduces the time spent understanding and rewriting the code. Verification, data migration and training remain cost items in their own right. Our article on the cost of custom software for an SME details what makes a budget vary.

Be wary of promises of a complete overhaul in a few days. Even at Cisco, with unlimited AI resources, the team gave itself four weeks just to demonstrate that its method works.

If an internal tool is ageing in your company, show it to us. From Rumilly, we look at the code and the data with you, then tell you frankly whether it is better to modernise it, replace it or leave it alone for a few more years.

Frequently asked questions

What is legacy business software?

It is software still used every day but built on old technologies, often poorly documented and hard to evolve. It does the job, which is why it is kept, but every change costs more and more.

Can AI rewrite old business software on its own?

No. It speeds up reading the code, translating it to a recent technology and writing tests. A developer must still validate every change, and the business rules must be confirmed with the people who use the tool.

Can software be modernised without interrupting the business?

Yes, that is the point of a module-by-module migration. The old and new systems run side by side, and each switchover happens after the data has been compared.

How long does modernising business software take?

It depends on the size of the tool and the state of the data. With a gradual migration, a first module can be delivered without waiting for the end of the project, which avoids a long period with no improvements.

Business softwareModernisationArtificial intelligenceLegacy softwareCustom software

How much would your project cost?

Project type, features, timeline: a costed estimate in two minutes, with no commitment.

Estimate my budget

Is your business software getting old?

Show us the tool your business runs on: we will tell you frankly whether to modernise it, replace it or keep it as it is. Write to us or book 15 minutes with Jérémy.