AI consultant for small & mid-sized businesses (MKB)

AI voor het MKB: where it actually pays back for a Dutch SME

Most articles about AI for SMEs in the Netherlands describe what large enterprises do, then tell you to copy it at a smaller scale. That advice rarely survives contact with a real mkb-bedrijf, where there is no data team, the IT budget is committed for the year, and the people who would run an AI tool are already running the company. A useful AI consultant MKB engagement starts from those constraints rather than ignoring them — which is why our first deliverable is an audit, not a build.

For the Dutch MKB, AI earns its keep in a handful of recognisable places. The first is repetitive document and text work: reading invoices, contracts, support tickets, CVs or insurance claims and turning them into structured data your existing systems can use. The second is forecasting and planning — demand, inventory, staffing, cash flow — where a model trained on your own history beats a spreadsheet rule of thumb. The third is search and retrieval: letting staff (or customers) ask a question in plain language and get an answer grounded in your own manuals, policies or product catalogue. None of these require you to "become an AI company." They require one bounded system, wired into the tools you already own. Across AI mkb-bedrijven we have worked with, the pattern that pays back is the narrow one: pick the single task that costs the most hours, prove the model on real data, then ship it into the workflow where those hours are spent.

Sector patterns we see across Dutch SMEs

• E-commerce & retail: product-description generation, demand forecasting to cut dead stock, and support automation that drafts replies in Dutch and English.
• Logistics & transport: route and capacity forecasting, document extraction from CMR/freight paperwork, and exception flagging before a delay becomes a complaint.
• Healthcare & care providers: administrative load reduction — intake summarisation, coding support and reporting — handled inside AVG-compliant boundaries.
• Finance, accountancy & professional services: reconciliation, anomaly detection, and retrieval over regulation and internal policy so staff stop hunting through PDFs.
• Marketing & web agencies: we act as the AI engineering partner behind the work you ship to your own clients — see generative AI — because Crux Digits is an engineering consultancy, not a competing marketing or web agency.

What the €2,500 AI audit actually examines

The fixed-price AI Audit & Strategy step exists so you never spend €20,000 proving the wrong thing. It is a short, focused engagement that produces a ranked list of opportunities with an honest read on payback and risk for each one. We look at four things in turn.

Data, process, integration and people

• Your data: what you already collect, where it lives, how clean it is, and whether there is enough of it to train or ground a model. Most AI projects that stall, stall here — so we check it first.
• The process behind the use case: how the task is done today, how often, by whom, and what a 30–50% time saving would actually be worth in euros.
• Integration reality: the systems a solution must talk to — your ERP, webshop, CRM, ticketing or accounting stack — and how hard each connection is.
• The people: who owns the workflow, who would use the tool day to day, and what change it asks of them. A technically perfect model nobody adopts saves nothing.

You leave the audit with a prioritised shortlist, a recommended first project, and a clear go/no-go — including the cases where our honest answer is "AI is not the cheapest fix for this yet." That candour is the point of paying a senior consultant rather than a sales team.

Betaalbare AI without cutting the corners that matter

"Affordable" should describe the price, not the quality of what you get. The way we make AI affordable for the MKB is by sequencing spend, not by skimping on engineering. The fixed-step pricing ladder — €2,500 audit, then a €20,000 proof of concept, then production from €50,000 — means each stage is small enough to approve on its own and earns the right to the next. You can stop after the audit. You can stop after the PoC. You are never locked into a six-figure commitment to find out whether the idea works.

Three levers that lower the real cost

• Scope discipline: one bounded use case beats a broad transformation programme for a business your size, and it costs a fraction as much.
• Subsidy: custom AI development usually qualifies for the Dutch WBSO R&D tax credit, which reduces the net cost of a PoC or production build — sometimes alongside MIT or SLIM funding.
• Ownership: because the solution ends up belonging to you, there is no perpetual licence quietly compounding for years after delivery.

For comparison, day-rate guidance sits around €150 per hour, but the reason we lead with fixed prices is precisely so you are not exposed to open-ended hourly billing on a project whose scope is still being discovered.

Boutique vs enterprise vs the "weekend AI" agencies

There are broadly three places a Dutch SME can buy AI help, and the trade-offs are real. The large consultancies — Xebia, Xomnia, Capgemini and peers — bring genuine depth, but they are built for enterprise transformation: long discovery phases, layered teams, and budgets that assume a corporate balance sheet. At the other end sit web and marketing agencies that rebranded around "AI" the moment the hype arrived, wiring a chatbot to a third-party API and calling it a strategy. Neither is wrong for everyone; both are a poor fit for a serious mkb-bedrijf that wants something it can rely on and afford.

What boutique means in practice

• Senior people stay on the project. The person who scopes your work is the person who builds it — there is no handover from a partner to a graduate.
• You own the result. Code, model and documentation are handed over so you are not dependent on us to keep the lights on.
• The track record is real: 13 delivered case studies across computer vision, NLP, forecasting and healthcare. Client names stay confidential, but the engineering is genuine — read more about Crux Digits and our approach to AI consulting in the Netherlands.

Compliance-first: EU AI Act and AVG built in from day one

For a Dutch SME, AI is not only a technology decision; it is a compliance one. The EU AI Act is now phasing in across the bloc, and it classifies systems by risk — meaning the obligations on an AI tool depend on what it does, not on how big your company is. A small firm using AI to screen job applicants or support a credit decision can fall into the high-risk category just as an enterprise would. The penalty for getting that wrong is not theoretical, and "we didn't realise the rules applied to us" is not a defence.

We design with that in mind rather than bolting it on later. That means data minimisation and a lawful basis under the AVG (the Dutch implementation of the GDPR), clarity on where data is processed and stored, documentation of how a model reaches its outputs, and a human kept in the loop wherever a decision affects a person. For most MKB use cases the right answer is a system that keeps sensitive data inside boundaries you control — important for healthcare, finance and HR applications in particular.

Practical questions we settle before any build

• Which EU AI Act risk tier does this use case fall into, and what does that require?
• What personal data does the system touch, and is there a lawful basis and a data-processing agreement to match?
• Where does inference happen, and does any data leave the EU?
• How will you explain, log and audit the system's decisions if a customer, employee or regulator asks?

From proof of concept to a production system you own

The gap between a demo that impresses in a meeting and a system that runs every day is where most AI projects quietly die. A proof of concept proves the idea works on your data; production is where it has to keep working under real load, with messy inputs, staff turnover and the occasional bad day for the underlying model. Closing that gap is engineering work, and it is what the from €50,000 production step buys.

What "production" includes

• Integration into your live systems, so the output lands where work actually happens — not in a separate dashboard nobody opens.
• Guardrails and monitoring that catch drift, failures and edge cases before your customers do.
• Handover and ownership: the documentation and knowledge transfer that let your own team operate, extend and eventually take full control of the solution.

That ownership principle is the throughline of every engagement. We are a boutique partner, not a dependency you cannot leave. Across our work — generative AI, machine learning, computer vision, data engineering and LLM optimisation — the goal is the same: a working AI capability that belongs to your business, sized and priced for the MKB, and built to satisfy both your finance director and a future EU AI Act auditor. If you want a sense of the range, the case studies show the breadth, and the pricing page shows exactly what each step costs before you commit a single euro.