Why the future of quant investing may depend less on AI models, and more on the framework surrounding them

Artificial intelligence is rapidly reshaping the investment industry. From machine learning to generative AI, asset managers are increasingly integrating advanced technologies into research, portfolio construction and risk management. Yet according to Mike Chen, Head of Next Gen Quant at Robeco based on Boston, the industry risks placing too much emphasis on the models themselves while underestimating where the real long-term competitive advantage may emerge.

“The model alone is no longer the edge,” Chen says. 

In his view, AI should not be regarded as a disruptive break from traditional quantitative investing, but rather as the latest stage in its ongoing evolution. Traditional quant strategies were largely built around structured financial datasets and factor models such as value, quality and momentum. Next-generation quant investing expands that framework by incorporating alternative and unstructured datasets. Across the industry, these may include sources ranging from satellite imagery and geolocation data to textual sentiment and behavioral information.

The objective is not simply to analyze more data, but to answer investment questions that historically sat outside traditional quantitative frameworks. AI increasingly allows investors to build systematic proxies for softer variables such as management transparency, customer trust or employee engagement — areas that were previously difficult to quantify in a systematic way.

At the same time, Chen stresses that next-generation investing does not replace established factor investing. Traditional return drivers remain relevant, but AI and alternative data can provide additional layers of insight that complement these more established techniques.  
 

Portfolio Day 2026
Mike Chen will speak at Portfolio Day 2026 about “Using AI and New Data to Find Opportunities”. Join us at Portfolio Day on Thursday, 10 September. Register here.

Turning data into differentiated insight

A key theme throughout Chen’s thinking is that data alone does not create investment value. In fact, the growing abundance of information increases the risk of identifying statistically convincing but economically meaningless patterns.

“In a world of big data, if you search long enough you can find almost anything,” he says. “That is why you need a robust investment hypothesis before you ever look at the data.” 

For Chen, successful AI-driven investing still begins with economic intuition and investment logic. Machine learning and alternative datasets should validate a rational hypothesis rather than replace it. This remains true whether firms are analyzing employee sentiment, customer behavior or broader market dynamics.

He also argues that the most valuable alpha opportunities are unlikely to come from obvious applications of alternative data that are already widely adopted across the market. As more investors gain access to similar tools and datasets, straightforward signals are likely to become increasingly commoditized. The greater opportunity may therefore lie in second- or third-order insights created by combining datasets in more sophisticated ways. 

Why human oversight remains essential

Despite the rapid progress of AI, Chen repeatedly returns to the importance of human judgment and accountability. While AI can significantly improve efficiency, pattern recognition and information processing, it remains imperfect. Models can hallucinate, generate false confidence or reinforce user biases.

“You can outsource tasks to AI, but you cannot outsource accountability,” according to Chen. 

That principle is particularly important in institutional investing, where fiduciary responsibility, governance and client communication ultimately remain human responsibilities. According to Chen, the future of investment management will not be machine-only, but instead based on hybrid organizations in which humans and AI increasingly work alongside one another.

As AI takes over larger parts of the analytical workload, the role of investment professionals is likely to evolve towards oversight, judgment and understanding where these systems can fail. Investors may not need to build algorithms themselves, but they will need sufficient technological literacy to challenge outputs and recognize the limitations of AI-driven processes.

Infrastructure may become the real competitive advantage

Perhaps the central message from Chen is that the real edge in AI investing may no longer reside in the models themselves. As frontier AI models become more widely accessible, competitive differentiation is shifting towards the broader framework built around them.

For Robeco, that means focusing on infrastructure, governance, workflow design, data engineering and operational scalability. For Chen, the challenge is not simply to experiment with AI, but to embed it into repeatable, auditable and scalable investment workflows.

He compares these capabilities to the hidden foundation of a skyscraper: largely invisible, but essential to the stability of the entire structure. In practice, firms that focus only on the visible AI layer without investing sufficiently in data architecture, controls and governance may struggle to build repeatable and scalable investment capabilities. 

This also helps explain why scale increasingly matters in AI investing. Building institutional-grade AI frameworks requires ongoing investment in technology, engineering talent and operational infrastructure. At the same time, firms must remain flexible enough to adapt their investment processes as the technology evolves.

Conclusion: execution, not access, will define the winners

As AI becomes increasingly commoditized, access to powerful models alone is unlikely to generate durable alpha. The real differentiator will be how effectively firms combine technology with data, infrastructure, governance and human judgment.

For wholesale and institutional investors, that distinction matters. In the next phase of quant investing, competitive advantage may depend less on who has AI… and more on who can operationalize it in a disciplined, scalable and accountable way.