ERP and AI: 5 tips for reliable data in SMEs

Most SMEs already have an Enterprise Resource Planning (ERP), but few use all its functions. The result: re-entry, parallel Excel files and decisions based on incomplete information. During the webinar of the Luxembourg Digital Innovation Hub, L-DIH Talks "Getting more from your ERP: Mastering the basics" on 16 October 2026, several speakers detailed the basics to be mastered.

To extend the reflection, we interviewed one of the speakers, Joé Glodt, co-founder and CEO of MyPlus Solutions, which supports companies in Luxembourg and Germany in the automation of their ERP processes. Here are five concrete tips that emerged.

1. Measure the quality of your ERP data before automating

The first step is to look at the actual state of the data and check that all employees are working as the process intended. A common case: a salesperson in a hurry creates an address in a hurry to send an offer, and then the accounting department has to recheck everything at the time of the invoice.

Duplicates pose a particular problem. An article that has been entered ten times, and is already linked to offers or stock, cannot be corrected with a click. "The biggest obstacle for AI is data quality and poorly defined processes. If the process is not good, how do you expect AI to do its job well?" summarizes Glodt. Cleaning up existing data still requires a lot of manual work.

2. Limit the number of people who enter data

The more hands there are on the basic data, the more the discrepancies multiply. "If there are too many people entering the addresses, out of ten people, there are always two or three who say to themselves: 'It's not important, I'll do it quickly'," observes Joé Glodt.

So designate a few people responsible for each type of data (addresses, items, prices). They apply the same method each time, and the other colleagues forward the requests to them rather than creating cards themselves.

3. Review your ERP processes, not necessarily your software

Changing ERP is not an obligation. "You don't have to change software if it's good for the company, but you should never stop working on your processes," insists Joé Glodt.

A telling example: a company goes from ten to twenty people without ever adapting its data entry rules. The habits that used to work with ten people no longer hold with twenty. Take a moment to ask yourself three questions: how do we work today, do we enter too much or too little information, what would we gain from entering more?

The webinar offered a simple exercise to get started: draw on a page your real flow, from the prospect to the payment, by identifying each passage by Excel or email.

4. Introduce AI into your ERP in stages, with human validation

The AI can already read an email, extract information from an enquiry and prepare an offer. But it doesn't decide on its own. MyPlus recommends a two-step progression:

• the user uploads an email into the tool themselves, receives a pre-filled form, and then corrects what needs to be corrected;
• once trust is gained, the tool processes the inbox in batches and checks the data upstream, but a person validates before any entry into the ERP.

Human control is still essential, for example to check prices and delivery dates on an order confirmation. Another condition is that the ERP must be open to external tools, which is not always the case.

Mr. Glodt cites a company that processes nearly 200 offers per day: the work used to involve two people, but now it employs one and a half. To fill an empty ERP, he estimates that it takes about 6 minutes to manually enter an item, compared to 40 seconds with AI. This is a significant time saving. 

5. Involve teams from the start

The fear of seeing AI "lose jobs" comes up among almost all customers. The answer lies in involvement: one contact person per company, who then relays the information internally. Employees test a trial version, identify missing information or steps, and then propose their own use cases. Nothing is imposed.

The challenge is also to reallocate time for more useful tasks. In sales, according to Joe Glodt, we now spend almost more time in front of the ERP than in contact with customers.

Where do I start?

It all depends on your starting point:

• You are still working on Excel: list your key processes and the data they use.
• You have a closed ERP: check the possibilities of integration with external tools before investing.
• You have an open ERP: audit the quality of your master data and identify a first AI use case.

In any case, the first step to improving your processes is to know your existing data. 

Are you an SME looking for a service provider to support you in your digitalization? The L-DIH Marketplace is the place to find it. We invite you to visit this free space bringing together local solution providers.