AI in ERP Sounds Simple on a Slide. It's Anything But in Production
Why AI in enterprise ERP is harder than the vendor demo suggests — and what building Kompas in-house at ZentixSoft taught us about it.
Agnessa Tomashevska
CEO at ZentixSoft
Recent news about adding AI to SAP and similar enterprise systems always sounds clean: plug in a model, automate reporting, boost efficiency. On a vendor's slide, it looks like a matter of weeks. In reality, enterprise companies get stuck on this for months — and the reason is rarely technical.
Why "just plug in AI" doesn't work in practice.
The first reason is customization. A ready-made ERP system, even one with AI built in, is designed around an average business process. But no real business runs on an average process. For the AI functionality to actually work — not just demo well on test data — the company still ends up paying for customization, just now with an AI layer on top. The second reason is legacy data: AI is only as good as the data it runs on, and cleaning up years of accumulated, inconsistent data is the longest, least glamorous part of the project. The third is team resistance — people don't rush to trust a system that suddenly "decides on its own."
We went through this ourselves — which is exactly why we built Kompas. When we were looking for a system to manage sales and internal processes at ZentixSoft, we tried the ready-made options — Monday, HubSpot. And every time we ran into the same paradox: for the system to actually cover our needs, it had to be customized almost 100%. Essentially, you pay for a SaaS subscription — and then you pay again to turn it into something that actually fits your business. The hardest part turned out not to be the automation itself, but picking the right system in the first place. So we took a different path and built Kompas — our own system that grew out of our real, often disconnected processes: sales, lead generation, HR bots that we built piece by piece until they came together into one coherent system.
Kompas is a good fit for businesses with a similar pain profile: B2B companies with multichannel or long sales cycles, product companies, teams that work with tenders, businesses where CRM is closely tied to a partner network.
A sober view.
Not every company needs AI functionality in its ERP just because it's trendy. Three questions worth asking:
- Is our data actually ready for AI?
- Is our process stable enough to automate?
- Is the team ready to accept a system that takes on part of the decision-making?
