Building for Reliability: AI in Regulated Investment Advice

Markus, co-founder of a new financial advisory app, opened his talk with a direct question for the room of builders. "Who of you have a portfolio of stocks that you have invested or ETFs?"

The question grounded the room in practical stakes. His company is tackling the challenge of providing regulated investment advice for individual stocks, a space where trust is non-negotiable. He gave an example of a user with a portfolio of five stocks. The app would provide advice on each stock, suggesting whether to increase, sell, or decrease a position, and even recommend new stocks that fit the user’s profile. He contrasted this with the standard disclaimers on finance YouTube videos, which explicitly state they are not professional financial advice. His product, he explained, is exactly that: regulated financial advisory for retail customers.

Advice Has to Be Deterministic

The core challenge is a fundamental mismatch between AI and regulation. "AI is probabilistic," Markus explained. Financial advice, however, "must be deterministic." A compliant product cannot give different answers to the same inputs. An investor cannot rely on a system that suggests buying a stock one day and selling it the next if the underlying data has not changed. He described this inconsistency as "noise or some fluctuation" that is not a reliable basis for financial decisions. For a regulated entity, any system that is not repeatable and auditable will "not survive any regulating" scrutiny. Markus's work is a case study in resolving the core tension of applied AI, proving a deterministic, regulated product can be built on top of a probabilistic technology.

Reliability Is the Product

The team’s answer is to reframe what they are selling. "We are selling the reliability for regulated investment advisory," Markus clarified. He challenged the builders in the room to consider the value their own AI overlays provide to users. For his company, the overlay's value is consistency and auditability.

This is what he called "the boring part," a process refined over three years of research and development. They use rule sets to ensure the advisory output is consistent and can be audited. This systematic approach is what allows them to work toward a BaFin license, the approval from Germany's financial regulatory authority. Markus described the EU as their target market. "The license is not from AWS," he noted, contrasting a cloud provider’s services with a financial regulator’s legal approval. In a world where AI has made intelligence cheap, he is banking on the idea that "reliability is the thing that's still expensive."

Human Experts, Not Adversarial Agents

During the Q&A, an audience member asked if the company used adversarial agents, where different AIs compete to find a better answer. Markus’s response was clear. "That's not how we are doing that," he said. "We have actual human experts who are financial people who are doing this research."

While the app removes the human advisor from the direct customer interaction to provide a scalable service, it keeps human experts at the core of its research process. When a user asked for clarification on personalization, Markus explained that the service mirrors the one-on-one sessions his co-founder used to provide, analyzing a user’s specific portfolio and profile to offer advice with a clear rationale behind it. But unlike other services, the user keeps their assets with their own broker. The app provides the advisory layer. He mentioned testing the system with a persona from Denmark and receiving tailored recommendations for Danish stocks, which proved how personalized the output could be. For every piece of advice a user receives, "there's actually a human experience behind it."

The Ask: B2C Go-To-Market

After two years focused on building this reliability layer, the company is close to its launch. They are about to finalize the regulatory process and receive the BaFin license. Now, the team's focus is shifting from engineering to distribution. "We are now focusing on to go to market," Markus shared.

He turned the end of his talk into a request for knowledge from the room. "So, if anyone has some experiences regarding a startup product that's to utilize and be to see, happy to connect and learn about that." It was a frank admission that building a regulated, deterministic engine is only half the battle. The next chapter is about finding and winning over customers. He invited people to sign up for a waiting list to be part of the initial test user group, closing on the concrete next step in turning a reliable product into a trusted one.

---

SQ Collective hosts Coworking Fridays for founders, operators, and AI builders working through real product questions in Singapore.

Join an upcoming Coworking Friday: https://lu.ma/ai-labs

Explore SQ Collective: https://www.sq-collective.com

Michael