Ai in Finance

This is a special edition episode recorded live at the UT CATT 2024 Global Analytics Conference on AI in Finance. Brad moderates a panel with three guests: Tal Schwartz, CEO and founder of AI Funds; Suhir Holla, CEO and founder of MyStockDNA; and Kyle Wiggs, CEO and founder of UX Wealth Partners (the show's sponsor). All four panelists have direct experience building and deploying AI in the investment management space, and Brad notes they could have easily gone five or six hours.

What AI Actually Does in Portfolio Management

The panel moved quickly past the buzzword stage into specifics. Tal Schwartz at AI Funds uses machine learning to process alternative data sets and generate investment signals. He explained that the value proposition isn't replacing human analysts but processing data volumes that no human team could handle: structured and unstructured data, satellite imagery, transaction data, social sentiment, all feeding into models that identify patterns for portfolio construction.

Suhir Holla approaches it from the investor's side with MyStockDNA. His platform uses AI to analyze a portfolio's actual underlying characteristics, essentially the DNA of your holdings. Most investors have no idea how much overlap exists in their portfolios, or how correlated their supposedly diversified positions really are. MyStockDNA maps those hidden relationships. Suhir described situations where an investor thinks they're diversified across ten funds but is actually concentrated in the same handful of risk factors.

Kyle Wiggs is building AI directly into UX Wealth Partners' TAMP infrastructure. The firm offers an AI-driven model marketplace where advisors can access quantitative strategies that were previously the domain of institutional investors. A solo RIA can now implement strategies with the same sophistication as a billion-dollar allocator. Kyle emphasized that the technology is mature enough to be production-grade, not experimental.

The Emotional Risk Frontier

The most compelling concept from the discussion came from Suhir: the "emotional risk frontier." Traditional portfolio theory builds efficient frontiers based on risk and return metrics. But for real investors, the binding constraint isn't the Sharpe ratio or standard deviation. It's the point at which they panic and blow up their own portfolio.

Suhir posed the question directly: can AI customize portfolios to individual emotional risk tolerances? If Brad panics at a 20% drawdown but another investor panics at 10%, their portfolios should be built differently, not just in terms of expected volatility but in terms of the probability of hitting each person's specific emotional breaking point. The panel agreed this could be where AI makes the single biggest impact on real-world investment outcomes. Not by generating alpha, but by preventing the behavioral destruction of wealth that happens when investors sell at the bottom. In other words, the best portfolio is one your client will actually stick with.

Beyond Markowitz: Regime-Adaptive Models

The panel directly challenged the traditional Markowitz efficient frontier framework that has dominated portfolio theory for decades. The core problem: correlations between asset classes aren't stable. They change based on economic regimes, and they spike during crises, which is exactly when diversification matters most. The 60/40 portfolio, long treated as gospel, showed its limitations in 2022 when stocks and bonds fell together. AI systems that adapt to changing correlation structures offer a meaningful upgrade over static models that assume fixed relationships between asset classes.

The discussion also covered personalization at scale. Instead of offering three to five model portfolios based on a generic risk questionnaire, AI could enable advisors to construct thousands of individualized portfolios that adapt in real time to each client's specific circumstances and risk tolerances. Kyle noted that UX Wealth Partners is building toward this vision, using machine learning to create dynamically customized solutions for each advisor's client base. Tal added that the challenge isn't the AI itself but the data infrastructure and regulatory framework needed to support it at production scale. The models work today. The plumbing needs to catch up.

Professor Kumar, who was also present at the conference, added academic grounding to the panel's observations. The discussion touched on how the entire field of portfolio theory needs updating to account for the computational power now available. Mean-variance optimization was revolutionary in the 1950s when calculations were done by hand. Today, machine learning can evaluate portfolio combinations numbering more than grains of sand on Earth, optimizing for far more sophisticated measures of risk and return than the simple Sharpe ratio. The panelists agreed that the theoretical frameworks taught in finance programs are decades behind what practitioners can now implement.

Key Takeaways

• AI Funds processes alternative data (satellite imagery, transactions, social data) through machine learning to generate investment signals that no human team could produce manually at scale.
• MyStockDNA reveals the hidden overlap and correlation in portfolios that investors don't realize they have, mapping the actual "DNA" beneath the ticker symbols.
• The "emotional risk frontier" concept proposes building portfolios to individual behavioral breaking points, not just financial risk metrics. The best portfolio is one your client actually sticks with.
• UX Wealth Partners is building AI-driven model marketplaces that give solo RIAs access to institutional-grade quantitative strategies through their TAMP platform.
• The panel challenged the static Markowitz framework and 60/40 default, arguing AI-driven regime-adaptive models better handle the reality that correlations change and spike during crises.

Listen to the full conversation on Spotify, Apple Podcasts, or YouTube.