TL;DR
- In 2015, I told AMweb that the profession was becoming less accessible because nobody could start without a diploma anymore.
- I can now substantiate that complaint: 29.7% fewer passed Wft exams in six years, based on my own adviser-shortage analysis.
- AI helps train new advisers, but it does not solve the shortage of qualified people.
- Two smaller observations from 2015 still hold: the bonus-malus system and "customer-centricity" as an empty slogan.
- I was completely wrong about blockchain insurance, and that is shaping how I assess AI hype today.
What I said about the diploma shortage
On August 18, 2015, I answered seven questions for AMweb's summer interview series. Theo van Vugt asked what I would tell then Dutch finance minister Dijsselbloem. I said the profession was becoming less accessible because people could no longer start without a diploma.
"Our profession has become less accessible because of this," was my exact line. At the time, that was not a dataset. It was a business owner's observation from day-to-day practice.
I also argued for more room in simple non-life products. Under supervision of an experienced manager, a junior employee could responsibly advise on straightforward coverages. Complex products such as mortgages obviously require deeper qualification.
The numbers I can now place next to it
This summer, I lined up six years of CDFD exam data. Between 2019/20 and 2024/25, the number of passed Wft exams dropped by 29.7%, from roughly 17,800 to roughly 12,500. Eight out of nine modules now score lower than five years ago.
Those 12,500 passed exams are not 12,500 new advisers. A professional qualification usually requires multiple modules. The cleanest intake proxy is Wft Basic. In 2024/25, around 5,600 people passed it, down from 7,300 five years earlier: 24% fewer.
Based on that intake, the current adviser age profile and expected retirement outflow, I built a cohort model through 2040. The full calculation is in adviser shortage: the 2040 generation is already born.
Demography adds a second decline. The cohorts that should supply successors are smaller, while the average age of PE candidates rose by four years in six years. Anyone working in this profession in 2040 is already born, and that group is simply smaller.
Follow me on LinkedIn
AI helps, but does not solve the shortage
In my own company, I already use AI to train new staff, including NotebookLM on Wft learning material. It clearly speeds up learning. AI also adds value in file preparation and first-draft advisory memos.
But AI does not increase the number of valid diplomas. The AFM requires a human with the right licence to advise clients. That is why my earlier point still stands: your licence is your right to exist.
Recruitment alone does not solve this either. If I hire an adviser away from another office, national capacity does not grow. The real lever is customer value per qualified adviser.
There is also a newer risk: if AI takes over all junior work, practical training disappears with it. I described that earlier as never-skilling.
Two smaller points that still hold up
The bonus-malus system. In 2015, I called this my first reform priority because most customer confusion came from this area. Eleven years later, that is still true in my day-to-day experience.
"Customer-centricity" as an empty phrase. I wrote this in 2015 too. The slogan is everywhere, but the execution is still often shallow.
Where I was completely wrong
In that interview, I predicted blockchain insurance would disrupt the industry in the near term. I even expected a market outsider to make the first serious move.
Eleven years later, I do not see that breakthrough. As far as I can tell, no meaningful blockchain insurer has emerged in the Netherlands, and many initiatives never moved beyond pilot phase.
In hindsight, I was asking the wrong question. I focused on what technology could theoretically do, not on a concrete customer problem that people actually felt.
What a failed prediction teaches me about AI hype
The difference between my correct diploma call and my wrong blockchain call is mainly the source. The diploma claim came from daily operational pain. The blockchain claim came from theory I had not seen in production myself.
That distinction now guides how I assess AI claims. I strongly back applications I build or test myself. Claims that come only from keynotes or LinkedIn posts get much more caution.
I am not writing this to pat myself on the back for one point and excuse myself on the other. I am writing it because this contrast shows exactly which predictions are worth taking seriously.
The 2023 birth cohort in the Netherlands counted 164,000 children. That means the shortage I described in 2015 is now largely locked in. In eleven years, I will review my next prediction the same way.
Sources
* The original 2015 interview: amweb.nl (opens in new window) * My adviser-shortage calculation: marcdiks.nl (opens in new window) * The blockchain article I referenced in 2015: amweb.nl (opens in new window)