Can AI help investors with questions (not just answers)?
By Jan Szilagyi on August 13, 2025
A lot of the buzz around LLMs right now is centered on their impressive ability to handle increasingly complex questions with apparent ease. And let’s be clear: this is no small feat. You’re talking about AI understanding a user’s question, figuring out how to answer it, gathering the necessary data, running the analysis, and then providing an answer. All of this might have taken hours of valuable time away from the user that could be better spent doing something else, like coming up with another question. And this is where things get interesting.
At Reflexivity, our goal of driving the cost of investor curiosity down to zero isn’t just a nice-to-have. It’s a game changer. Investors are grateful that the platform handles a lot of the heavy lifting: pricing a one-touch EURBRL digital option while also building an independent forecast for NVDA, all with minimal effort. Our proprietary Knowledge Graph stitched together otherwise entirely separate data sources, and linked them in a way that made data easily discoverable and analyzable.
But here’s the kicker. There’s an even bigger need that portfolio managers and analysts are eager to solve: building the curiosity muscle. They don’t just want answers; they want better questions. AI is still largely thought to be clueless when it comes to independent, creative thinking. It doesn’t seem to be capable of the kind of original thought that has long been the domain of the human brain, with all its mysterious, messy ways. But what if we are wrong and AI can help with that too?
This is the frontier we’re aiming to conquer with Reflexivity’s unprompted AI. We’re already helping investors answer their questions. Could we take it a step further? What if we could help them identify the questions they should be asking, but aren’t?
Let’s start with the “known unknowns.” You know, those gaps in knowledge we’re aware of but haven’t filled yet. If we ask the right questions, we can turn those gaps into “known knowns.”
However, we also have the “unknown unknowns” in Donald Rumsfeld’s parlance. These are the gaps in our knowledge we don’t even know exist. This is the stuff that in hindsight appears obvious but no one saw coming. The black swans. The 5-sigma events. And that’s where the real challenge - and the real opportunity - lies.
The dynamic Knowledge Graph is key to achieving this. It provides Reflexivity with a basic mapping of investing relationships that matter, and gives it “situational awareness.” Whenever it detects a dislocation in US 10 year yields, it will know to look for ripple effects in utility stocks or the term structure of commodity curves.
We have also given it the freedom to uncover new ones, initially relying on some degree of pattern recognition and relatively elementary statistics. And guess what? It is starting to turn from a passive participant into an active investigator that is able to highlight potentially overlooked ideas. It will uncover links between companies that may defy conventional categorization, and adjust its revenue expectations for FedEx or UPS based on logistics commentary from Amazon.
This will change the paradigm. Rather than prompting Reflexivity with “Tell me what factors are driving a company’s discount to its peers” the user will instead be prompted:
“I found another company that is just as exposed to the Supply Chain Automation chain theme but trades at a substantial discount due to very temporary factors.”
We are just getting started but the promise of this endeavor is absolutely immense.