Hiring ICT specialists became easier between 2022 and 2024, and the large Dutch and Belgian organisations that are moving fastest on AI are not choosing between hiring, training and partnering. They do all three. In our analysis of Eurostat data, 80.2% of Dutch companies with 250 or more staff ran ICT with their own people and external suppliers together in 2024, up from 68.1% in 2022.
This piece is for the IT manager, CIO or information manager at a 250 to 5,000 FTE organisation who has to staff an AI programme in 2027 and explain the plan to a board. Crux Digits is the AI partner that works alongside your ICT partner or internal IT team, so we have a view on the split, and we state it below with the numbers behind it.
Is it still hard to hire ICT specialists for an AI programme?
Yes, but less than in 2022: 69.8% of large Dutch ICT recruiters had a hard-to-fill vacancy in 2024, down from 79.7%.
That 9.9-point drop is the good news in the data. Across the EU, the share of large recruiting firms with hard-to-fill ICT vacancies fell in 22 of 27 member states, to 68.2%. Belgium fell further, from 86.4% to 75.6%. Germany stayed tight at 80.3%.
The easing did not come from lower demand. In 2024, 64.2% of large Dutch firms recruited or tried to recruit ICT specialists, third in the EU and up from 58.3% in 2022. The Netherlands also has the fourth-largest ICT workforce in the EU: 708,800 people in 2025, 7.2% of employment. The market is deeper and busier than ever, and it is working better.
Eurostat counts ICT specialists, not AI specialists, so treat this as the nearest measured signal for AI roles rather than a direct one. The full study sets out the tables and their limits.
Why will hiring alone not staff an AI programme in 2027?
Because the constraint is qualified people, not salary: 63.5% of large Dutch recruiters cited too few applicants, 45.7% cited pay.
The four reasons Dutch firms gave in 2024 were a lack of applications (63.5%), applicants without the relevant ICT qualifications (60.2%), applicants without relevant work experience (50.2%) and salary expectations (45.7%). Pay ranks last in the Netherlands, and lower than in Germany (59.1%), Belgium (52.7%) or the EU as a whole (51.2%).
That changes the arithmetic of a hiring plan. Raising the offer helps less in the Netherlands than in Germany, because the shortfall is people with the right skills and experience. Two levers address that directly: growing those skills inside your own organisation, and bringing in experience from a partner while you do. A bigger national ICT workforce does not solve it on its own either. Across the 27 member states the correlation between ICT workforce size and hiring difficulty is 0.17, which means much of the lever sits inside the organisation.
How do large Dutch and Belgian firms staff ICT today?
With a mix: 80.2% of large Dutch firms and 85.7% of large Belgian firms combine their own ICT staff with external suppliers.
The share of large Dutch firms that run ICT with their own employees alone halved in two years, from 17.9% to 9.2%. In Belgium it is 6.6%, the lowest in the EU. Own people did not disappear: at 89.3% of large Dutch firms, own employees perform ICT functions. What changed is that suppliers sit next to them as a normal part of the operating model.
Training grew alongside. In 2024, 67.5% of large Dutch firms trained staff who are not ICT specialists in ICT skills, nearly double the 34.5% of 2014. Belgium leads the EU on this measure at 83.9%.
For an AI programme this is a useful starting position. Your organisation already knows how to work with suppliers, contract them and audit them. Adding a specialist AI partner extends a model you run today; it does not invent a new one.
When should you hire, train or partner for a role in an AI programme?
One rule per seat: hire if the role outlives go-live, train if the knowledge belongs to a process, partner for a build peak.
Applied to a typical programme, the rule sorts the seats like this. The product owner for AI, the data engineer who keeps pipelines healthy after launch and the architect who decides how AI fits your landscape are permanent roles: hire or appoint them. The process owners in finance, HR, customer contact or operations who define what good output looks like, and the service desk staff who field the first questions, need AI literacy more than a new job title: train them. The first build, the evaluation set, the integration with your systems and the audit documentation are a peak of specialist work that a partner can carry while your own team learns: contract it, with the hand-over written in.
This is the reverse of how many programmes start. They hire for the peak and then look for work for the new hires once the build is done. The data supports doing it the other way round.
What should a CIO do in the next twelve months?
Five moves, each tied to a finding from our analysis of Eurostat data, cover the next year.
Move 1: put recruitment lead time in the plan. With a seven-in-ten chance (69.8%) of at least one hard-to-fill ICT vacancy at a large Dutch recruiter, and three in four (75.6%) in Belgium, the permanent seats should be opened first, before the build starts.
Move 2: hire for experience and qualifications, not for a pay premium. Since applicants and skills are the constraint and salary the least cited reason, widen the pool instead: internal ICT staff who want to move into data and AI work, with a training budget attached.
Move 3: train the business side early. Large Dutch firms already doubled ICT training for non-specialists in ten years. AI literacy for process owners fits the same budget line, and the AI Act now asks for it: Article 4, as amended by the Digital Omnibus on AI, requires providers and deployers to take measures to support the AI literacy of their staff.
Move 4: contract a partner for build capacity with a hand-over clause. The mixed model is now the norm at 80.2% of large Dutch firms. Write into the contract that repositories, documentation, evaluation sets and runbooks transfer to your team, so the knowledge stays when the partner leaves.
Move 5: write the seam down. With your ICT partner, your own team and an AI partner all involved, decide per system who changes it, who monitors it and who gets called when it misbehaves. One page per system is enough, and it is the kind of record an auditor asks to see.
What does 2027 look like if the trend holds?
If the 2022 to 2024 direction holds, hiring keeps easing and the mixed model becomes the default for AI, as for ICT.
That is an outlook, not a forecast. The study contains no projection, and the ICT skills questions run every two years, so the next reading will show whether the easing held. What the existing data does show is a steady direction: a larger ICT share of employment in 2025 than in 2015 in all 27 member states, nearly twice the share of large Dutch firms training non-specialists compared with 2014, and about half as many running ICT alone as in 2022.
For an AI programme that is a positive picture. The skills market is deeper than it has ever been, firms that train their own people are no longer the exception, and organisations have learned to run suppliers side by side. The programmes that do well in 2027 will be the ones that decided early which seats they own and which they borrow.
Where does Crux Digits fit in a hire, train or partner plan?
Crux Digits fills the partner seat for AI build work, next to your team and ICT partner, and hands over what it builds.
We are not an ICT partner. We do not run infrastructure, a helpdesk or hosting, and we assume you already have someone who does. We design and build AI systems and the software around them, document them for audit, and hand over code your organisation owns, so the permanent roles you hire inherit a working system and its paperwork. The ICT staffing study has the full league tables behind every number used here.
For how we scope an AI programme with your own team and your ICT partner, see applied AI and machine learning for mid-sized organisations.
Last updated: 28 October 2026.
A consultant tells you where AI pays off; Crux Digits also builds it. A fixed price per step, one named expert, from Utrecht.
AI consultancy in the Netherlands →
