If you are buying business software with AI inside it in 2027, the question is not who demos best but who leaves you a system your IT organisation can run, test and change. The eight Dutch firms below build that kind of software, by their own published descriptions. Crux Digits is our pick for a first, bounded AI system on your own data; the other firms below are the better choice for multi-year programmes, managed operations or consumer products.
Disclosure: Crux Digits publishes this list and is listed first. We say why, and who the others are better for. The list is written for IT managers, IT directors, CIOs and information managers at organisations of 250 to 5,000 employees who already have an ICT partner and want to add AI-powered software without adding a system nobody can support.
What counts as AI-powered business software?
AI-powered business software is a working application in your own environment where a model makes or prepares a decision inside a business process. Think of an invoice that is read and coded before it reaches AFAS or Exact, a werkbon that is checked against the bestek, a support question answered from your own documentation with the source cited, or a demand forecast feeding the planning screen of your ERP.
What it is not: a strategy deck, a proof of concept that only runs on a laptop, a Copilot licence rollout, or a chatbot pasted onto a website with no connection to your systems. None of those is software you own.
For the broader market of custom software firms, with or without AI, our existing guide to the best custom software development companies in the Netherlands covers large consultancies, maatwerk specialists, product studios and nearshore teams. This page narrows that field to firms that put AI into production software.
How did we choose these firms?
A firm is listed only if its own website shows it builds software with AI inside, not only AI advice. We checked every site on 29 October 2026, and each entry links to the page we verified.
- Builds, not only advises. The firm's own pages describe building AI applications, data and AI products, or AI functionality inside business software.
- Dutch delivery. An office in the Netherlands and work for Dutch organisations.
- Published specifics. Named services, named cases, certifications or prices on its own site. We state prices only where the firm publishes them, and none of the others publishes a price for AI software work.
- An honest fit. For each firm we name the buyer for whom it is the better choice than Crux Digits.
The list is not a ranking beyond our own pick. Crux Digits is listed first because we publish this page and say why below; every entry after ours is in alphabetical order. There are no paid placements, affiliate links or reciprocal-listing deals. We left out firms whose sites describe AI advice or training only, and firms whose AI claims we could not find on their own pages.
Which firms build AI-powered business software in the Netherlands in 2027?
Crux Digits, Avisi, Betabit, Info Support, Kabisa, Q42, Xebia and Xomnia each describe building production software with AI inside on their own sites.
1. Crux Digits (our pick for a first AI system on your own data)
Disclosure: Crux Digits publishes this list and is listed first. We say why, and who the others are better for.
Crux Digits builds custom software with AI at the core: web applications, document and process automation, grounded assistants on your own knowledge, machine learning on your own data, and the integrations with an ERP, CRM, WMS or AFAS that make them useful. It is based in the Utrecht region and works in English and Dutch.
Why it is first, in checkable terms. The published prices are fixed per step, all excl. VAT: an AI audit at €2,500 in one to two weeks, a Production-ready MVP at €20,000 in four to six weeks, and a production launch from €50,000 in eight to twelve weeks. Scope and price are fixed in writing before work starts. At handover you receive the code, the model configuration, the prompts and the test set, so your own team or another supplier can run and change the system.
Crux Digits is an AI partner, not an ICT partner. It does not take over hosting, the service desk or systems management. It works next to the supplier you already have, builds inside your identity provider and the interfaces you already maintain, and hands over a runbook your second line can read.
Better for someone else if: you need a forty-person programme, 24/7 managed operations, or a consumer app where design carries the product. The firms below are built for those.
2. Avisi
Avisi builds business-critical software and has more than 25 years of experience. Its AI offer includes AI-assisted development, a Solution Studio session that aims to produce a working prototype in about four hours, proofs of concept on its own AI platform Phie, and sector work in energy, finance, trade associations and the installation sector. Avisi states ISO 27001, ISO 27701, ISO 27017 and ISO 27018 certification and a SOC 2 Type II report.
Best for: an organisation whose procurement asks for a certified supplier that also runs a managed cloud platform, especially in energy, finance or installation.
3. Betabit
Betabit calls itself a specialist in managed software development: it builds, tests and maintains mission-critical maatwerk software, end to end or by strengthening your own development team. It has nearly 200 people across Rotterdam, Amsterdam, Apeldoorn and Eindhoven, names AI as one of its three strengths, and is a Microsoft Solutions Partner for Digital App & Innovation and for Data & AI. One published case describes an AI application for Rotterdam Harbour Holding built within a month.
Best for: a Microsoft and Azure estate that wants one supplier to build and keep developing a mission-critical application with AI in it for years.
4. Info Support
Info Support positions itself around critical IT built on craftsmanship and strengthened with AI across the lifecycle. Its AI page splits the work into agentic business (customer assistants on internal knowledge, intelligent document processing, forecasting, process automation, copilots inside business applications) and agentic engineering (AI-assisted coding, testing and documentation). It also lists a sovereign AI platform and states that its services are hosted entirely in the Netherlands.
Best for: an organisation in agrifood, energy, finance, ports and logistics, industry, mobility, government or healthcare that wants build, management and training from one supplier.
5. Kabisa
Kabisa in Weert and Amsterdam sees software development as craftsmanship and builds custom software through its Software Artisans, with a Data & AI practice that applies generative AI to business processes alongside data engineering, data platforms and BI. It lists managed services, ISO 9001 and ISO 27001, and a software partnership with ASML since 2014.
Best for: a technology or manufacturing organisation that wants a long-running software partner who also manages what it builds.
6. Q42
Q42 is a digital product studio in Amsterdam and Rotterdam with 120+ engineers and 25 years of experience. It lists AI applications among its products, with examples such as a ChatGPT-based kiosk interface for KLM, and holds ISO 27001:2022.
Best for: a customer-facing app or platform where product design and scale matter as much as the AI inside it.
7. Xebia
Xebia describes itself as consulting, engineering and training, with AI-first as its stated direction. Its AI services include agentic AI, GenAI solution development and AI managed services, next to AI-native digital engineering and cloud and data modernisation. It has a presence in 16 countries and its Dutch offices are in Hilversum, Amsterdam and Eindhoven.
Best for: an enterprise programme that needs several teams, cloud and data modernisation and training for your own engineers under one contract.
8. Xomnia
Xomnia in Amsterdam works under the line "We make AI work". It builds data and AI products from A to Z with MLOps, GenAI and analytics engineering, builds AI agents, and builds cloud data platforms on Azure, AWS and Polars.
Best for: an organisation whose AI case depends first on a data platform and machine learning engineering, with a data team that will own the result.
How do you tell production AI software from a demo?
Ask for seven artefacts before you sign; a firm that builds production AI software can name each one and show where it ends up in your repository. This is the test we would apply to every firm on this list, ourselves included.
- An evaluation set: real cases from your process with the answer your own experts agree on, kept apart from anything used to build the system.
- An acceptance metric and a baseline, agreed before the build, so "it works" means a number.
- Logging of inputs, outputs and model version, with a retention period that matches your privacy policy.
- A fallback path: what happens when the model is unsure, wrong or unavailable, and which human receives the case.
- Pinned model and prompt versions, stored in your repository, plus a documented way to swap the model provider.
- Integration through interfaces you already maintain, behind your identity provider, rather than a parallel data copy.
- A runbook: what to check weekly, the thresholds, and who to call when the output is wrong but nothing is down.
Press hardest on points 1, 4 and 7. They are what lets your team judge the system a year later, without the builder in the room.
Who is the provider under the EU AI Act when a firm builds AI software for you?
Often you are: the AI Act defines a provider as the party that develops an AI system or has one developed and puts it into service under its own name. That wording sits in Article 3(3) of Regulation (EU) 2024/1689. If a software firm builds a system to your specification and you run it under your organisation's name, the provider duties can land on you, not on the builder.
That changes what you should ask a builder for. For systems that talk to people, Article 50 requires providers to design them so that people are informed they are interacting with an AI system. Article 25 adds that a deployer who puts its own name on a high-risk system, substantially modifies it or changes its intended purpose so that it becomes high-risk is treated as its provider.
For stand-alone high-risk systems under Annex III, such as AI in recruitment or credit scoring, the AI omnibus that the Council approved on 29 June 2026 moved the application date to 2 December 2027, as summarised by Lewis Silkin. For 2027 procurement, put three lines in the contract: who is the provider, who keeps the technical documentation, and who updates it when the model changes. You can check a system's likely tier with our EU AI Act risk checker.
What security questions should you ask an AI software builder?
Use the OWASP Top 10 for LLM Applications 2025 as the agenda: it is OWASP's public list of risks specific to software with a language model inside. The OWASP list opens with LLM01 prompt injection and includes LLM02 sensitive information disclosure, LLM06 excessive agency and LLM10 unbounded consumption.
Turn those four into questions your security officer can score. How does the system treat text in an uploaded document or email that tries to give it instructions? Which data can the model see, and can it repeat that data to a user who should not see it? Which actions can the system take on its own in your ERP or CRM, and which need a human click? What caps the cost and volume of model calls if something loops? Good answers are design decisions and test cases.
Also ask how the firm uses AI in its own coding. The 2024 DORA report found that AI adoption had a negative effect on software delivery stability and throughput, so AI-assisted development is a reason to ask for test evidence, not a reason to expect a lower bill.
Which firm fits which kind of project?
Match the firm to the shape of the work: one bounded system on your data, a long-running managed application, an enterprise programme, a data platform, or a consumer product.
- One process, one system, on your own data, priced before work starts, handed to your IT team: Crux Digits.
- Mission-critical maatwerk that a supplier keeps developing for years, on Microsoft: Betabit.
- Critical IT with build, management, training and a sovereign AI platform from one supplier: Info Support.
- A certified supplier with its own AI platform and managed cloud, in energy, finance or installation: Avisi.
- A long-running software and data partner that also manages, in technology and manufacturing: Kabisa.
- An enterprise programme with many teams and training: Xebia.
- A data platform and machine learning engineering first: Xomnia.
- A consumer-facing product where design and scale lead: Q42.
Where does Crux Digits fit?
Crux Digits fits an IT organisation that wants one AI system on its own data, built inside its standards and supportable by its own people or ICT partner. The Production-ready MVP takes four to six weeks. We take one process from audit to production at a fixed price per step, we work next to your ICT partner rather than replacing it, and we do not sell hosting, a service desk or 24/7 operations.
If your project is a multi-year programme, a platform rebuild or a consumer app, one of the other seven firms on this page is the better call, and we will say so in the first conversation.
See how we build AI-powered business software on custom software development.
Last updated: 29 October 2026.
Software built around your process and data, with AI where it changes the outcome. Something working every week, and the code is yours.
Custom software development in the Netherlands →
