InsureBench
About InsureBench
An independent initiative measuring how well AI models perform on the Dutch Wft-Basis insurance-advisor exam.
InsureBench is not a product of any employer, regulator or publisher. Results are only published after a full run has completed and each run is verified by the author before publication.
No account required, no cookies, no tracking. Public pages serve aggregated scores only — individual answers and the question bank remain private.
Who builds this
MD
Marc Diks, independent researcher on AI in financial services.
Conflicts of interest and governance
The author also works in a commercial role at a Dutch insurance intermediary. To prevent conflicts of interest, the following safeguards apply:
- InsureBench runs on a private domain and private infrastructure, and uses no data or systems from the author's employer.
- InsureBench does not publish model recommendations used in production by the author's employer. Models are selected based on market availability, not internal choices.
- All published results are traceable to the runs in the public CSV and JSON. There is no selective publication.
Changes to this governance are recorded in the changelog.