Is this work of equal value?
Paste two job descriptions. We score each one on its own against the four factors of Directive (EU) 2023/970 and show where they meet and where they diverge.
Once the law is in force, any employee can ask what colleagues doing work of equal value earn on average, split by gender. To answer that, you first have to know which roles are of equal value.
Your collective agreement settles the pay structure, not the categories
A collective agreement fixes scales and a pay structure, and can meet the requirement of a pay structure for equal and equivalent work. That is worth something, and it is not a part you have to rebuild.
Alongside it sits a second, separate obligation: establishing categories of workers performing equal or equivalent work. That one stays with the employer. A collective-agreement scale is not a category: two roles with different names can be of equal value, and workers in the same scale do not automatically belong together. One category can even span different collective agreements.
That is where the risk sits too. If a pay-transparency obligation has not been met, it is on you to show there was no discrimination when a claim is brought. With a recorded weighing per role you stand on the other side of that same rule. This check shows what such a weighing looks like, for two roles.
Art. 4 lid 4 Richtlijn (EU) 2023/970
"een categorie van werknemers (...) op niet-willekeurige wijze gegroepeerd op basis van (...) niet-discriminerende en objectieve genderneutrale criteria"
Wetsvoorstel 36949, Memorie van Toelichting (implementatie Richtlijn (EU) 2023/970) (MvT par. 4.3). Laatst geverifieerd 28 juni 2026.
How this weighting works
Three choices that separate a verdict you can explain from one where all you can say is that the computer thought so.
On content, not on title
The job title never reaches the model. Titles carry status and gender load: "executive secretary" and "office manager" score differently on the title, not on the work. Personal names and pronouns are stripped too.
Each role scored separately
Role A never knows role B exists. That keeps the weighting from anchoring on surface similarity, and every score traces back to a quote from the description itself.
The verdict is arithmetic
The comparison happens in code, not in the language model. Two roles with the same total but a very different profile therefore do not get an automatic green light.
The four factors
Skills
Level of knowledge, Experience, Complexity of problems, Social and communication skills
Effort
Mental effort, Emotional effort, Physical effort
Responsibility
For results and decisions, For people, For resources or information, For external relationships
Working conditions
Physical conditions, Psychosocial conditions, Irregular hours and availability
The directive prescribes no weighting. The four factor groups therefore carry equal weight, a quarter each. Equal weighting is the most neutral choice in the absence of a rule.