WBSO for AI development: what qualifies, what doesn't
WBSO reimburses 36% of S&O payroll costs for technically new development (50% for starters), up to EUR 391,020 per year. Only genuine technical R&D counts: applying or prompting existing AI tools doesn't qualify; building a new algorithm or model yourself does.
WBSO is separate from the SLIM program. SLIM reimburses learning to work with AI; WBSO reimburses developing something technically new yourself. Most Dutch SMEs adopting AI fall under SLIM. WBSO is for the smaller number of companies with their own technical staff doing genuine R&D: building a new model, developing a new algorithm, working in technically uncertain territory.
That distinction is the point of this page. Presenting an application project as R&D is the fastest route to rejection.
What is WBSO?
What: The S&O-afdrachtvermindering (WBSO) reduces the payroll tax a company remits for staff performing technical research and development work.
Amount: 36% on the first EUR 391,020 of R&D payroll costs per year (50% for starters), 16% above that.
Sole proprietors without payroll staff: a fixed S&O deduction of EUR 15,979 in 2026, plus EUR 7,996 extra for starters, in place of the payroll-tax reduction.
Who: Companies with R&D staff on payroll (or sole proprietors) doing technically new work.
Applying: Rolling intake, at least 1 calendar month before your desired R&D period starts.
The 36%/50%/16% percentages and the EUR 391,020 threshold are the confirmed 2026 parameters (RVO, updated via the 2026 Tax Plan). They're revised annually; check rvo.nl for the current year before you build a dossier.
Does AI development qualify for WBSO?
RVO tests for technical novelty, not functional novelty. RVO's test comes down to this: the work must exceed your company's own existing technical knowledge and skill, and the research must not be routine in nature. Without technical uncertainty upfront (not knowing whether your approach will work), there's no R&D in the WBSO sense.
Generally qualifies
Developing a technically new algorithm, your own model, or software that solves a concrete technical problem you and your R&D staff have to figure out how to solve yourselves. The recurring question: did the technical approach already exist inside your company, or did it have to be developed from scratch?
Generally doesn't qualify
Applying existing offerings: calling a language model via an API, configuring an AI tool, prompt-engineering a chatbot, or fitting an existing AI service into a workflow. That may be functionally new for your business, even valuable, but it isn't technical R&D. Routine work without technical risk (building a website, setting up a shop, software configuration) also falls outside the scheme.
How the application works
WBSO doesn't have fixed windows like SLIM. You can submit a new application every calendar month, as long as it lands at least one calendar month before your desired R&D period. Want to start January 1? December 20 of the prior year is the deadline.
Once granted, there's a binding obligation: hour registration. Which R&D staff member, which day, which project, how many hours, logged within 10 working days of the work being done. At the end of the calendar year, you report the actual realized hours and costs to RVO through the mededeling, due March 31 of the following year. Only hours from staff on your own payroll (or yourself as a sole proprietor) count; outsourced R&D work doesn't.
Why software applications get rejected
Three reasons come up most often in advisory practice around WBSO for software.
- Insufficient technical justification. The project description explains what the software should do (functional), not what's technically uncertain about it upfront. RVO assesses the latter.
- Applying instead of developing. The project, in practice, consists of configuring or wiring together existing AI services, without developing anything technically new yourself.
- Work that isn't on payroll. R&D outsourced to a freelancer or agency doesn't count; WBSO only applies to hours from your own R&D staff.
Where this page can't answer for you
The honest version of this page also says what it can't judge.
Whether your specific project is technically new enough
That's an RVO determination, made per application, based on your company's history and your R&D staff's background. No page and no advisor can guarantee that in advance.
The exact S&O hourly rate for your calculation
RVO uses either a fixed hourly rate or your actual hourly wage, depending on the choice you make in the application. The calculation method is on rvo.nl; this page doesn't run the numbers for you.
Whether the percentages hold in 2027
The 36%/50%/16% percentages and the EUR 391,020 threshold are set annually through the Tax Plan. Nothing here is a guarantee for next year.
WBSO combined with SLIM/MIT
These programs rarely conflict, as long as costs aren't double-claimed. SLIM reimburses learning to work with AI tools (workshops, internal playbooks); see SLIM Round 2. WBSO reimburses developing something technically new yourself. MIT (Feasibility or R&D) is a third option for larger innovation projects, often with a knowledge institution, and runs its own application round per province. Most Dutch SMEs adopting AI for their own workflows fall under SLIM; WBSO is for the smaller number building something technically new themselves.
Where Orellis fits
Orellis is set up to check, before you invest time in a dossier, whether your AI project clears the WBSO technical-novelty bar. Where it does, Orellis drafts the technical project description RVO needs: what's technically uncertain, what you don't know upfront, and why applying an existing model isn't enough. No guarantee of approval. What you get is a file that answers the right question, and where WBSO isn't the fit, an honest answer that SLIM or MIT fits better.
We review everything that ships. This page was drafted with our own AI stack and reviewed by a human before it went live.
Don't want to sort this out yourself?
Tell us where your team's time goes: five questions, about 90 seconds. We read every one and reply within 2 business days, and flag which programs fit your case: SLIM, WBSO, or neither.
Tell us where the time goes SLIM Round 2 all AI subsidies