Kerris ICT is an AI company based in Vaals, South Limburg. I build applications that take over a concrete task: the phone, documents, recurring messages.
AI is only useful once it takes over a task that costs time today.
I start with the process, not the model: which steps are repetitive, where does it go wrong, and what may be automated? Then I build the application: a phone assistant, a document parser, or an integration that replaces a manual step.
Which task costs the most today and is the most predictable? That's where it starts.
Where may the data live, what may the model see, who is responsible for what?
A working version on real input, not on sample data.
What goes well, what goes wrong, and where should a person take over?
Projects I actually built, not sample cases.
Healthcare organizations spend a lot of time on phone handling: simple questions block the line for emergencies.
AI-driven phone assistant with speech recognition, LLM triaging and seamless handover to human staff.
Automatic handling of routine questions, healthcare staff relieved, patients helped faster.
Orders with customer data and print PDFs came in at Boogaard Textiles. Those orders had to be placed into the print program with cutting lines, but it was so much manual work that the €100k cutting machine stayed unused. Instead, everything was cut with scissors.
A Laravel API that automated the entire process: process incoming orders, generate cutting lines, place them paper-efficiently in the print program. A dashboard tracked everything with barcode scanning. No more lost orders. Connected to the website, financial administration and postal dispatch.
The cutting machine is finally in daily use. A permanent employee was hired to operate it. The boss's son was freed from manual cutting work, editing cutting lines and order management. The workflow went from a day to 20 minutes.
Government organizations need to match CVs against CompetentNL competencies, a manual, error-prone process. HR staff lost a lot of time interpreting widely varying CV formats and mapping them to standardized competencies.
A Laravel + React application that automatically parses CVs, extracts competencies and matches them against CompetentNL standards. The tool recognizes free-text experience and translates it into structured competency profiles. Fully bilingual English/Dutch, with Docker deployment.
Runs live against the official CompetentNL SPARQL endpoint using the official CompetentNL→ESCO mapping (120 mapping rules), not a home-grown AI approach. Built for organisations that have to match CVs against CompetentNL.
Sectors where this work pays off most.
Background on this service, without the sales pitch.
What clients usually want to know before they call.
Then it hands over to a person. That's not a failure path but the design: the application handles the predictable questions and passes the rest to someone who does know.
There are conditions. Since the AI Act, a caller must be able to tell they're talking to a system rather than a person. I build that in; the background is in the knowledge base.
Storage (transcripts, recordings, documents, appointments) sits on your own server or on infrastructure under your own management, with an agreed retention period and documented access. The processing itself (speech and language model) runs via specialised vendors under a data processing agreement, with training on your data contractually excluded; which vendor that is and in which region they process, I name explicitly per project. If you want everything inside a single environment, that is a custom build, and we discuss what it costs. No voice cloning and no reselling of call data.
No. It takes away the repeat questions so there's time left for the conversations that need attention. On a busy day that means fewer interruptions, not fewer people.
Custom web applications, API development and full-stack platforms. I build scalable solutions with modern technologies.
More infoTechnical advice, architecture engineering and digital transformation. For companies that want to make better technical decisions.
More infoI work from Vaals, on site across South Limburg and beyond.
Got an idea, a problem or a project? I'm happy to help. No obligations, just a good conversation.