Amsterdam Data AcademyHow the AI Skill Monitor measures
What we count, where the data comes from, how often we measure and what the numbers can and cannot tell you. Written for journalists, researchers and anyone who wants to quote the data.
Source
All job postings that are live in the Netherlands and Belgium on the measurement day, as collected by a large job-board aggregator that combines postings from employer sites, job boards and recruitment agencies. One posting is counted once, regardless of how many sites carry it.
Frequency
One measurement per month, in the first working week. Each measurement is a snapshot of all postings live on that day. The first measurement dates from September 2026; every month adds a data point.
Unit
Postings, not people or hires. We report the number of postings that mention a skill, and the same number per 1,000 postings, so that a busier or quieter job market does not distort comparisons over time.
What we count
Hard AI skills. Each skill is counted with one fixed English key phrase. Using the same phrase every month keeps months and countries comparable; the phrase is chosen as the most specific term employers use for that skill.
| Skill | Counted with |
|---|---|
| AI governance and compliance | AI governance |
| AI literacy | AI literacy |
| AI security | prompt injection |
| Building AI agents | AI agents |
| Classic data skills | machine learning |
| Context engineering and RAG | retrieval augmented generation |
| Data quality and governance | data governance |
| Evaluating and testing AI | model evaluation |
| Process automation with AI | workflow automation |
| Working with LLMs and prompting | prompt engineering |
Soft skills. Measured only within postings that mention AI. They show what employers expect from people who work with AI, next to the technology itself.
| Soft skill | Counted with |
|---|---|
| AI-assisted decision making | AI-assisted decision making |
| Adaptability and learning | continuous learning |
| Change management | change management |
| Critical review of AI output | critical thinking |
| Ethical judgement on AI | AI ethics |
| Stakeholder management | stakeholder management |
| Translating tech to business | translate technical |
| Working alongside AI systems | human-AI collaboration |
AI roles. Counted on the job title, with a small set of common title variants per role.
| Role | Counted with |
|---|---|
| AI Coach | AI Coach, AI Strategist, AI Adoption Lead |
| AI Engineer | AI Engineer, ML Engineer, Machine Learning Engineer |
| AI Governance Officer | AI Governance Officer, AI Ethics Officer |
| AI Product Owner | AI Product Owner, AI Product Manager |
| AI Trainer / Consultant | AI Trainer, AI Consultant |
| Automation Engineer | Automation Engineer, RPA Developer |
| Chief AI Officer | Chief AI Officer, CAIO |
| Data Engineer | Data Engineer, Analytics Engineer, BI Engineer |
| Data Scientist / Analyst | Data Scientist, Data Analyst |
| Prompt Engineer | Prompt Engineer |
Tools. Postings that name the tool anywhere in the text.
| Tool | Counted with |
|---|---|
| Azure OpenAI | Azure OpenAI |
| BigQuery | BigQuery |
| Databricks | Databricks |
| GitHub Copilot | GitHub Copilot |
| LangChain | LangChain |
| Microsoft Fabric | Microsoft Fabric |
| Power BI | Power BI |
| Snowflake | Snowflake |
Sectors and countries
Every measurement covers the Netherlands and Belgium separately, for all sectors together and for ten sectors: Accounting and finance, Consultancy, Creative and design, Education, HR and recruitment, Healthcare, IT and software, Legal, Marketing and communications, Sales and customer service. Sectors follow the job category the employer selects for the posting.
Long-term trend and outlook
For the trend since 2019 we use the AI tracker of Indeed Hiring Lab (CC BY 4.0): the share of all Dutch job postings that mention AI, as a monthly average. The outlook is a linear regression over the last 12 months, projected one month ahead, with a 90% prediction interval. The per-skill outlook uses our own monthly measurements; with fewer than four measurements it is damped and capped, and shown as a direction rather than a forecast.
Limitations
- A posting is a signal of demand, not a hire. One vacancy can stay online for weeks; others are filled without a public posting.
- English key phrases miss postings written fully in Dutch or French without the English term. This mostly lowers the counts for non-technical roles and sectors; the trend over time remains comparable.
- Small numbers move fast. In sectors with few postings a change of a handful of postings can look like a large percentage. We show the base for every number.
The yearly research
The monitor is one half of the picture. Every year we also run a survey among organisations and in-depth interviews with HR, L&D and business leaders, published in The Skills of Tomorrow's AI Workforce, in partnership with research agency Dialogic. The 2026 edition is based on a panel of 80 L&D experts.
Using the data
You are welcome to quote and reuse the figures with a reference to the source and a link to this site. For background, data on request or an interview, book a call with our team.
