Run the Test: Three Ways Kelly Works Differently Than a Typical LLM
How to run the test
Ask Kelly a question you already know the answer to. Not because it's hard, but because you can check the answer yourself. Ask an AI tool something you don't already know and you have no way to tell whether it's giving you the real answer or just a plausible-sounding one.
"What's the best pair trade for Home Depot?" is a good one to start with, since you already know the answer is Lowe's. Ask a general AI assistant this and it will likely say Lowe's too, confidently, because it has read enough retail research to know the two compete for the same customer. Ask Kelly and you might get the same name, you might not, since it depends on context. What's different is how Kelly gets there, what data it uses and what it does next, which is how and why the two answers may differ. Let’s use this one question as the walkthrough to see how Kelly and typical LLMs differ.
1. Kelly AI agent is systematic and deterministic, not a best guess
A pair trade suggested in isolation or in the context of a broader portfolio can lead to different answers. Kelly doesn't answer "Lowe's" simply because it reads like the obvious comp. It screens candidates for the closest factor exposures to Home Depot and/or the overall portfolio, depending on the context, so the systematic risk cancels, but then also tests something the risk model only assumes rather than proves: whether each peer's residual, stock-specific returns actually move with Home Depot's once that shared factor risk is stripped out. Kelly would rank Lowe's first because its idiosyncratic returns show the highest realized correlation with Home Depot's over the trailing months; a tighter, more mean-reverting spread than any other candidate, not just because it's the obvious sector peer.
That's what "computed, not inferred" means in practice. A general LLM predicts the next likely word. Kelly routes the question to a real calculation on Omega Point's Investment Intelligence engine instead. Ask about the same stock on the same day and you get the same numbers every time. What differs firm to firm is never the math, it's the analysis built around it. Kelly also knows your portfolio in detail and hence has the context to then be able to make appropriate suggestions that are actually relevant and actionable.
2. It's grounded in real information, traced to source
Kelly's answer isn't pulled from the general internet or limited to what a model reads before its training cutoff like a traditional LLM. It comes from Omega Point's Investment Intelligence engine, using established and trusted third party and propietary data sources. This is the same intelligence engine the world's largest investors have trusted their analytics with for over a decade, and every number in the result traces back to the data and model that produced it. You can ask Kelly “show me your sources” and it will show you exactly how it got there.
It also uses the same governance and SOCII compliance for Kelly that its used for 13 years, so you only see what you're entitled to see. If you can't see it, Kelly can't see it either. Nothing here is a plausible-sounding sentence stitched together from patterns. It's a number you can trace, computed the same way every time you ask.
3. Kelly anticipates the next question, including ones you haven't thought to ask
Kelly doesn't stop at "Lowe's." It surfaces the next few candidates too, ranked the same way, before anyone asks, along with why each ranks where it does. A PM's real next question is rarely "is that all", it's "what's my next best option, and why does it rank below Lowe's." Kelly answers that unprompted, and sometimes the question you hadn't thought to ask yet, because knowing what a 25-year market veteran would want to know next is built into how Kelly was designed.
A typical LLM and Kelly can both tell you Lowe's is the pair trade for Home Depot. Only one of them can show why, in numbers, and already has the next few ranked behind it.
That's the pattern across everything Kelly does, not just pair trades. Ask it the way you'd ask a colleague, then ask the follow-up, and the question after that. Go into Kelly and try it: "What's the best pair trade for Home Depot in my portfolio?" Don’t have Kelly, sign up here.
