Latest / Stock Picking with Fexingo: Individual Equities, Research, and Building a Concentrated Portfolio / How AI Agents Are Reshaping Stock Picking
Transcript
- Lucas: So there's a fascinating story out today that I think gets to the heart of what stock picking actually is, or at least what it's becoming. CNBC reported that startups are building AI agents to trade 24 hours a day — not just execute trades, but do the research, read the filings, listen to earnings calls, and make decisions. Luna: I saw that. The idea of a fully autonomous AI stock picker is both exciting and a little unsettling. But how real is this today? Lucas: It's real enough that the headlines are here. The piece profiles a handful of firms that claim their agents can digest an earnings transcript, flag anomalies, and even generate a buy-sell signal faster than a human team. And this comes the same week Visa announces it's cutting 7% of its workforce — about 2,100 people — as part of an efficiency push driven in part by AI reshaping work. Luna: Right, Visa's move is a concrete example of AI replacing some human tasks. But is that directly comparable to stock picking? I mean, processing documents is one thing, but making a judgment call on a company's competitive moat? Lucas: That's exactly the question. Let's take Visa itself as a case study. A human analyst would look at Visa's network effects, its pricing power, the regulatory landscape, the shift to digital payments. An AI agent can process all that data too, but does it weigh the qualitative factors correctly? For instance, Visa's recent decision to cut jobs is a cost-saving move, but it also signals a cultural shift. Can an AI appreciate that? Luna: Probably not the way a seasoned analyst would. But the proponents argue that AI doesn't need to 'appreciate' — it just needs to find patterns that correlate with future returns. And with enough data, maybe it does that better than humans. Lucas: And there's evidence that ai driven funds have been beating the market in certain periods. But here's the thing: we're in a market now where stock picking is especially challenging. The S&P 500 is at 7,429, the Dow at 52,747, but small caps as measured by the Russell 2000 are down 0.4% in the last five days. There's a clear rotation happening. An AI might pick up on that rotation faster than a human, but it could also amplify momentum and create crowded trades. Luna: So you're saying AI agents could be great at short-term tactical moves, but for a concentrated portfolio of long-term holdings, the human element still matters? Lucas: Exactly. Building a concentrated portfolio means you have to have conviction. You're not spreading risk across fifty names; you're betting on ten or fifteen businesses. For that, you need to understand the business deeply — its culture, its management, its moat. An AI can give you a signal, but the final decision, I think, still rests on human judgment. Luna: But what if the AI is trained on decades of earnings calls and can detect subtle linguistic cues that humans miss? Some research shows that the tone of the CEO during a call predicts future stock returns better than earnings surprises. Lucas: That's a fair point. And there are firms now using natural language processing to score every word of an earnings transcript. So the line is blurring. But I'd argue that the best use of AI in stock picking is as a research assistant — not as the final decision-maker. Let the AI screen, summarize, and flag inconsistencies, but then the human applies context. Luna: That sounds like a hybrid approach. And given that Visa itself is using AI to cut costs, it's ironic that stock pickers using AI might be analyzing Visa's own ai driven efficiency. It's almost recursive. Lucas: Totally. And that's where things get interesting. As AI agents become more common, the market may become more efficient in the short term — less mispricing for humans to exploit. But I think there will always be opportunities in areas where data is scarce or qualitative, like turnaround stories or regulatory changes. Luna: So what does this mean for the individual stock picker listening today? Should they start using AI tools, or double down on fundamental research? Lucas: I think they should do both. There are free or low-cost tools that can analyze sentiment, track insider trades, even generate rough valuation models. Use them to save time, but don't outsource your thinking. The edge still comes from understanding something that the market hasn't fully priced in. Luna: And our listeners know we believe in the value of deep, thoughtful analysis. That's exactly why we keep this podcast ad-free — we want to provide honest, unfiltered perspectives without commercial interruption. If our conversations have helped you think about your portfolio more clearly, you can support us at buy me a coffee dot com slash fexingo. No pressure, just a way to keep the show independent. Lucas: Absolutely. And every bit of support helps us keep digging into these questions. So, getting back to AI agents: I think the key is to embrace the technology while remembering that a stock is not just a data point — it's a piece of a business. Luna: That's a good reminder. So, Lucas, do you think in five years the best stock pickers will primarily be programmers, or still fundamentally analysts? Lucas: I think the best will be analysts who can code enough to leverage AI, but who retain that deep business judgment. The machines will handle the grunt work, but the conviction — the willingness to make a big bet on a misunderstood company — that's still human. Luna: Well said. And with that, we'll leave our listeners to ponder that. Until next time.