The future of investment analysis: Q&A with Reflexivity CEO and Co-Founder Jan Szilagyi

Use cases for Generative AI (Gen AI) and agentic AI in financial services continue to grow on what seems like a daily basis. Companies, mainly startups, are offering solutions related to research, data and portfolio analysis that could transform how everyone from portfolio managers to wealth advisors do their jobs.

The startup Reflexivity has designed a suite of advanced AI-powered products specifically designed for institutional investment workflows, putting its technology to work helping clients with deep research, document intelligence and portfolio insights.

Jan Szilagyi, Reflexivity’s Chief Executive Officer and Co-Founder, formerly worked as a co-Chief Investment Officer at Swiss private bank Lombard Odier and portfolio manager at Fortress Investment Group. He characterizes Reflexivty as “the market platform I wished I had while managing a $15 billion global macro book.” He runs the company with a team of former portfolio managers and technologists.

Jelena Zec, Director, Citi Ventures, spoke with Jan in an email conversation about a host of topics, including the company’s 'self healing’ approach to data, how Reflexivity helps clients get the full benefit out of AI-powered tools and what issues keep him up at night.

Overview of Reflexivity

Jelena (Citi Ventures): How does Reflexivity approach the decision-making process in complex, high-stakes environments? In financial services, for example, there are a lot of variables and inputs, so what does your company do to either simplify things or empower decision makers?

Jan (Reflexivity): Reflexivity is a zero-hallucination, plug-and-play investment analysis platform. All data sources (Datastream, I/B/E/S, S&P Global etc.) are fully integrated and linked into the platform through proprietary Knowledge Graph. The Graph meticulously maps out critical relationships between assets and market, fundamental and macro drivers in real time, using both unstructured and structured data.

That means our analytics engine continuously scans for shifts and proactively notifies you when investment opportunities or risks to your portfolio emerge. To dig deeper, an LLM-powered interface allows for open-ended data queries with virtually unlimited analytical complexity.

A ‘Self Healing’ Approach to Data

Jelena (Citi Ventures): Can you tell us about your approach to AI self-reflection and introspection? How is this different from what others are doing and why does your approach set Reflexivity apart?

Jan (Reflexivity): We use AI self-reflection and introspection capabilities to create more robust analytical processes: Reflexivity is able to judge partial results and correct course (find better data sources, debug the code, clean or winsorize the data etc.) if it judges it is veering too far from the original plan. This self-healing means a slightly longer completion time but also a much more thorough and reliable analytical output.

Working Closely with Clients

Jelena (Citi Ventures): What are some roadblocks you have hit on your journey and how have you overcome them? What lessons could you share with others?

Jan (Reflexivity): The biggest obstacle we have faced so far is a relatively high degree of confusion around AI capabilities among our existing and potential clients. This has limited adoption – or confined it to only very trivial use cases – much like using a computer to perform Abacus-like tasks. So education and experimentation, a true thought partnership with a client, has been essential to overcome such reservations.

Empowering Humans to Tap into Troves of Data

Jelena (Citi Ventures): What role do humans play in deploying and using Reflexivity?

Jan (Reflexivity): A central one: a human user is the conductor of a vast array of analytical powers Reflexivity bestows on them. They are able to use Reflexivity’s awareness of key investing relationships - interest rates and bank stock prices, Iowa temperatures and harvest yields etc. - to easily discover and analyze a universe of data that was always available but remained un-discoverable and un-analyzable.

Wealthcare: Bringing Together Wealth and Investment Management

Jelena (Citi Ventures): Extending that point out, how do you see humans and AI systems interacting 5, 10 even 20 years out? What does that look like and what’s the evolution?

Jan (Reflexivity): The superpowers will only grow. AI - in investment management - is killing tasks, rather than jobs, and this will reward experienced users. A fully autonomous AI investment analyst is definitely within reach, but it’ll likely be in support of a human operator. Demand for “wealthcare” (wealth management/investment management) is virtually limitless (like with healthcare, you can’t have too much health or too much wealth).

As “wealthcare” gets better and more personalized, through support of intelligent systems like Reflexivity, it’ll simply offer a wider range of services to a wider range of people that previously weren’t served nearly as well.

Anticipating Users’ Queries and Needs

Jelena (Citi Ventures): How will you keep your technology and models relevant? What internal methods – data flywheel, proprietary training methods – do you/will you use? Any other insights there?

Jan (Reflexivity): Reflexivity is able to get better and better with use: the better it understands the nature of user’s questions, and the types of analysis asked of it (prompted AI), the better it’s able to actively highlight (unprompted AI) relevant market developments - risks and opportunities - that drive user engagement (back to prompted AI).

Keeping AI Focused on Customer Benefits

Jelena (Citi Ventures): What keeps you up at night?

Jan (Reflexivity): Currently, the evolution of AI is incredibly fast and exciting: but also unpredictable. As a developer of apps that leverage it, we have to stay on top of where its biggest strengths are and leverage it for the benefit of our customers. This list of capabilities is constantly shifting and requires a degree of informed guesswork to anticipate what it’ll look like a week, a month and three months from now.