“Thanks to AI, six of us can do the work of sixty analysts”

The East-Flemish fund manager Value Square has been working with AI to select stocks for about two years. “Due to stricter rules, analysts were barely following the small stocks. AI does help us do that.”

Koen Hoffman, CEO of the fund manager Value Square in Ghent, lets us browse an analysis report from a medium-sized listed company. The company's strengths and weaknesses are described in detail. The sector, the remuneration of top management, operational levers and the biggest financial risks are also extensively covered. It looks like the work of an analyst who has done a very thorough job. “But the report was rolled out of the computer in ten minutes,” says analyst Wouter Verlinden.

This is because it was created by AI. “We feed the AI tools with highly targeted data,” says Verlinden. “The annual reports, of course, but also the transcripts of the analyst calls, the presentations for analysts and investors, news articles, information about management, their background and how much they are paid in stocks.”

The result isn't perfect yet, he admits. “You have to see the result as the work of a junior analyst. There may be another mistake. So human control is always needed.” Verlinden estimates the error rate at 2 to 3 percent. “By feeding the model properly, we minimize the risk of hallucinations.”

Trial and error

Verlinden, who also has an IT background, started experimenting cautiously with AI two years ago. It quickly became clear that the possibilities should not be underestimated. “In stock research, AI allows for unprecedented productivity gains,” says CEO Hoffman. “We now have six of us doing the work of sixty analysts.” That's why Value Square also recruited an EY data specialist to help create reliable analysis models.

“That happened through a lot of trial and error. We've already changed our model ten times because new platforms became available so often,” says Kris Hermie, managing director and chief investment officer of Value Square. “Two people can't keep up with all the new tools that are being developed.”

“The big advantage of AI is that you can view many more companies quickly. For us, it's very important to be able to analyze more ourselves,” says Hermie. “Our focus has always been on small to medium-sized listed companies. Due to the stricter investment rules after the financial crisis, banks and stock exchange houses have significantly reduced the analysis of that type of stock. Some stocks are no longer followed, or only by two or three people,” says Hermie. “Then it's interesting to generate extra analysis capacity yourself.”

These analyses have evolved considerably over the past year. “The most important thing is to optimize the prompts, the commands you give to the language model,” says Hermie. “A good prompt explains Value Square's entire investment process to the AI model block by block, so that the resulting analysis perfectly shows what you're looking for in a company as an investor. Thanks to the integration into our own database, we can assess a company's capital investments against comments from peers. For example, the analysis is not limited to that specific company, but the analysis of the peers is immediately spun off.”

Chinese models

The AI analyses also put new companies on Value Square's radar. Ionos, for example, a German hosting company similar to Combell, which will hopefully do better after a few bad quarters, and Krones AG, a German 'hidden champion' in the beverage sector. “We make the final investment decision ourselves,” Hermie emphasizes, “but the basic analysis that AI draws up does help us.

Recently, Value Square gave a presentation to a number of family offices to show what the AI evolution means for analysts. The top five most used AI models for the applications they need include no less than three Chinese models, with Minimax and Kimi in places one and two. Google's Gemini 3 is in third place, Deepseek in fourth place. The language models were not developed to perform large calculations, so you should definitely pay attention to the free models, warns Hermie. “But the more advanced models do the same thing as people who get asked a difficult math question: they add a calculator.”

The question remains whether Value Square is now also achieving higher returns by using AI. A recent overview of the funds that invested best in Belgian stocks last year put Value Square second after Econopolis. The return of 20 percent was just below the performance of the Brussels stock market index Bel20.

‍

References

• Source: De Standaard
• Link: https://www.standaard.be/economie/beleggen-met-hulp-van-artificiele-intelligentie-we-doen-met-zes-het-werk-van-zestig-analisten/139451177.html
• Authors: Pascal Dendooven
• Photo: Fred Debrock
• Date: 17/03/2026

‍