Custom models for prediction, classification and automation — trained on your data, tuned to your domain, and built to earn their keep. When off-the-shelf AI can't see what matters in your business, a model built on your own data can.
Last updated: 11 June 2026
Generic AI is trained on the whole internet; your business is specific. Machine learning is how you capture what's specific — the patterns in your customers, your operations, your history — in a model that predicts, classifies or automates with an accuracy a generic tool can't match.
We focus on models that actually ship and pay back: scoped to a real decision, measured against a real baseline, and deployed where they make a difference.
Forecast demand, churn, risk or revenue from your own historical data.
Automatically sort, tag, prioritise or score records, leads and documents.
Surface the right product, content or next action for each individual user.
Take repetitive, judgement-based tasks off your team's plate, reliably.
Honest accuracy, bias and baseline testing so you know what you're really getting.
Models put into production and watched for drift — not left to gather dust in a notebook.
We turn a business question into a measurable modelling problem.
We assemble and clean the data the model will learn from.
We build, tune and honestly evaluate the model against a baseline.
We ship it into your workflow and monitor it over time.
Often less than people fear. We'll tell you early whether your data is enough for the problem — and if not, exactly what would close the gap.
We measure it against a clear baseline and report accuracy, errors and bias honestly. If it doesn't beat the baseline, we tell you rather than ship it.
Yes. Machine learning here means models trained on your data for prediction and classification; for language and content we also offer LLM optimisation.
Yes — we deploy models into your stack and monitor them, so they keep performing as your data and business change.
We test for it explicitly and, where relevant, design the model and process to keep decisions fair, explainable and defensible.
We build monitoring into every deployment — tracking key metrics and data distribution. We set alert thresholds upfront and include a retraining protocol so the model stays accurate over time.
We start with what you have and tell you honestly what gaps exist. In most projects we can train a working first model before data cleaning is complete — the roadmap shows you what to improve next.
Where the use case requires it, yes. For regulated or high-stakes decisions we build in explainability — SHAP values, feature importance, audit logs — so you can show how a decision was made.
We are Crux Digits, a boutique applied-AI firm in the Utrecht region working with clients across the Netherlands and the EU. Instead of a big consultancy, you get senior engineers hands-on from scoping to deployment, with no offshore hand-off. We build custom models trained on your own data, remotely or on-site, and keep the work aligned with the EU AI Act and GDPR from day one.
We begin with a short paid discovery to scope the model against a real decision, then quote a fixed price. A proof-of-concept can start from a few thousand euros; a production model typically from around 20k, larger builds more. Senior engineers do the work, you own the source code and IP, and the Dutch WBSO R&D scheme can offset part of the cost.
Find at-risk customers and the levers to keep them — so marketing targets the right cohorts instead of spraying spend.
Turn demand signals into a production plan, smooth mould-change losses and defend market position — for a €235M footwear manufacturer.
Learn the signatures of failure and flag them early — so maintenance happens on schedule, not in a crisis.
Tell us the question; we'll tell you honestly whether a model can answer it — in a free consultation.
Book a free consultation →