Asset managers have poured millions into developing proprietary AI systems, betting that custom-built technology will provide a competitive edge. But research from MIT suggests they may be making a costly mistake. Externally developed AI systems were twice as likely to reach full deployment as those built internally.
According to MIT's 2025 State of AI in Business report, which analyzed 300 public AI deployments and surveyed 153 leaders from 52 organizations, internally developed AI systems reach production at drastically lower rates than specialized external tools. In their sample, external partnerships with learning-capable tools reached deployment 67% of the time, compared to 33% for internally built tools.
"The core barrier to scaling is not infrastructure, regulation, or talent. It is learning," the report states. Most internally developed systems fail to retain feedback, adapt to context, or improve over time.
Organizations building proprietary systems often create "brittle workflows" – rigid implementations that can't adapt as business needs evolve. Most companies also lack the scale and specialization required to continuously refine AI models.
Stefan Klauser, CEO and Co-Founder of Aisot Technologies, an ETH spin-off working with European and U.S. asset managers, has observed this pattern. "Asset managers excel at understanding markets and constructing portfolios," he said. "But building AI systems that learn and adapt requires a different kind of expertise – one that typically sits outside traditional financial institutions."
Attempting to build AI entirely in-house is a bit like designing a Formula 1 car on your own. Achieving peak performance requires specialized engineering, rigorous testing, and continuous refinement. Partnering with experienced teams can provide that high-performance edge more efficiently.
This is exactly what Aisot Technologies delivers. Instead of generalist AI solutions, we focus on what asset managers value most: portfolio optimization, customization, and automation. By combining advanced machine learning with proven forecasting methods, we provide AI-driven analytics that generic tools simply can’t match.
We also tackle a critical challenge: AI hallucinations in financial forecasting. aisot’s time-boxed and fine-tuned language models train only on historical data up to specific points, eliminating the “look-ahead bias” that can affect general-purpose LLMs (learn more here).
Moreover, our solutions integrate seamlessly into existing workflows, requiring no major changes to how asset managers work. Most importantly, they are designed for continuous learning: the AI Insights Platform, for example, evolves through client feedback, adapting to real-world usage and growing more valuable over time.
This approach aligns directly with what executives prioritize: 66% want systems that learn from feedback, and 63% want systems that remember important information.
MIT’s research shows that differentiation comes not from using AI, but from how firms apply it to their unique needs. External partnerships offer a faster, more reliable path to value.
The most effective AI adopters no longer wait for perfect use cases or central approval. They drive adoption through distributed experimentation, vendor partnerships, and clear accountability.
The question facing asset managers isn't whether AI will reshape their industry – that's already happening. It's whether they'll spend years building systems with a high failure rate, or deploy external tools already in production.