Blog

Replacing a legacy application with AI: start with the business case

Replacing an old application is not an act of faith. Compare the full annual cost with the cost of a smaller, better-defined process.

Two colleagues looking at an older monitor together

A legacy application often stays in place longer than anyone would like. Not because people are happy with it, but because replacing it feels like a big job. There is a contract, there are screens people know, and there are exceptions that nobody ever wrote down properly. So renewing looks like the safe choice.

That safety is sometimes more expensive than it looks. The supplier's invoice is only part of the annual cost. Add the custom work needed for a small change, the time staff spend on workarounds and corrections, and the requests that get shelved because a change costs too much or takes too long. Only then do you see what the current process really costs.

Cost the process, not the whole IT estate

"Replace the application" is usually a poor first brief. An old application often holds many functions, while only some of them create work every day. Start with one well-defined business function. For example: checking incoming orders and getting them into the ERP system. Or answering customer questions with data from three existing systems.

A choice like that reduces the risk. The rest of the application can stay in place while the first bottleneck disappears. It also makes the comparison fairer. You compare the new version of one step with what that step costs and how long it takes today, instead of pitting a new digital world against an old one.

Set out the current annual cost in five lines:

current annual cost = contract + custom work + manual work + rework + deferred changes

For each line, enter an amount or time spent that you can back up. Do not use an estimate nobody can trace. If nobody keeps count of manual corrections, track them over a representative period of normal work. The result does not have to be perfect. It has to be something finance, operations and IT can discuss around one table.

A new solution has costs too

The other side of the business case is more than a build budget. Include integrations, access rights, testing, support, model usage and the cases where an employee still has to check the work. A solution with no licence fee is not automatically cheaper if it produces a lot of manual exceptions every month.

Keep one-off and recurring costs separate:

one-off = discovery + build + integrations + testing + go-live

annual = support + usage + exceptions + further development

Usage covers things like model calls, storage and document search. Exceptions are work that deliberately stays with a colleague. Those costs are the price of a boundary that stops an uncertain decision from going through automatically, not a weakness in the design.

The comparison is stronger when the new solution delivers one measurable change: less time between arrival and booking, less manual entry, less rework, or the ability to make a change in days instead of months. Decide in advance which of these counts and who owns the measurement.

Find the break-even point

A savings percentage without the assumptions behind it is worth little. A percentage only means something when it is clear which contract, which custom work and which process boundary were compared. A smaller replacement can pay off for an application with high maintenance costs and a lot of repetitive work. For a stable system with low volume, keeping it may be the wisest outcome.

So put three scenarios side by side: renew as things stand, renew with the change you need, and replace one process. Show what each scenario costs in the first year and after that. Also note which assumption has the biggest effect on the outcome. That is often the amount of manual work or the size of an integration.

The business case compares contract, manual work, rework and the cost of changes. The exceptions determine which work stays with employees. The target is work that is now copied between screens, waits for a change or has to be checked again every time. The old application does not always have to sit in the middle of that work.

Leave room for a no

A good discovery phase does not automatically lead to a build. It can also show that the source data needs fixing first, that the volume is too low, or that a standard change from the current supplier is cheaper. That is useful to know before a larger programme starts.

A precise savings figure only becomes credible when contract, process volume, process boundary and all assumptions are laid out side by side. Until then, the business case itself is the valuable part. It gives an executive grounds to stop treating a renewal as a given.

Three ways to start

Do you have an application whose bill goes up every year while the work stays the same?

Start small, with one process that comes up every week. Choose how: from half an hour behind the scenes to a session at your offices.

01

Look behind the scenes

Free

Half an hour, by invitation

See everything we build running in our sandbox, the Proeftuin. No sales pitch, unless you ask for one.

Request an invitation

02Recommended

Introductory call

Free

Half an hour on Teams

You tell us what you already do with AI. We tell you what fits, or whether it is still too early.

Call with Marcel Logtenberg

Book a call

03

Get started

From €495

At your offices, excluding VAT

An AI quick scan, an AI assessment or an AI assessment plus inspiration session, at a fixed price.

Get started