During the recent Christmas holidays, artificial intelligence infiltrated almost all my meetings, whether as a practical tool or as a topic of conversation. At one of those meals, surrounded by friends who are painting enthusiasts — and not exactly financial market aficionados — the discussion turned to the role of AI in the investment world. The group was an interesting mix: some of us work in finance, while others are investors. The debate arose immediately. I argued, with my usual enthusiasm, that AI is still far from replacing us in financial advising and investment management. The response was quick: “Do you really think you can analyze data better than AI?” I acknowledged that no human can process data at the speed of a machine. But the real challenge is not the computational ability, but how it is used. For AI to provide useful recommendations, it is essential to know what to ask; otherwise, it can lead us to poor decisions. Before returning to artistic topics, someone promised to send us a video about AI advancements, and I shared one of my usual experiments with these tools in my work. After decades in the industry, I consider myself capable of spotting their mistakes, without pretending to be more than someone with real experience. How have AI-managed funds performed? It is important to distinguish between funds that invest in AI companies and those managed by algorithms. In the U.S., several such funds were launched between 2017 and 2020, such as the Amplify AI Powered Equity ETF, the QRAFT AI-Enhanced US Large Cap ETF, and the QRAFT AI-Enhanced US Large Cap Momentum ETF (AIEQ, QRFT, and AMOM). Although they outperform the S&P 500 in certain periods, their performance over three and five years is lower than that of a traditional ETF on the index. For now, AI does not have a crystal ball better than the market. On December 23, I conducted a simple test: I asked ChatGPT, from my paid account, which stocks to invest in for 2026. Its response was strategic and cautious, suggesting sectors like critical infrastructure, essential energy, and health, with specific examples like Rheinmetall, Thales, Vinci, National Grid, Enagás, Roche, and Siemens Healthineers, while warning about risks and recommending funds or staggered entries. When I repeated the question without logging in, the tone changed: legal warnings, lack of personalization, and references to human analysis. In other words, AI referred to the opinions of human experts. I also tried Perplexity, DeepSeek, and Gemini. All responses were based on online searches, with few matches on specific stocks — only Rheinmetall was repeated — and all recommended caution and the use of funds or ETFs, although they ended up citing specific stocks. However, before giving recommendations, I would always ask the investor if they would need that money in 2026. If the answer is yes and they do not want to risk losing capital, perhaps short-term fixed income is the best option. Following generic advice without personalizing questions to AI can be risky. Does this mean that AI is not useful for investment advice? Not exactly. The value of AI depends on how and what it is asked, and the information provided to it. Additionally, one must consider the management of sensitive data and who has access to it. In fund management, it is worth remembering that even AI cannot predict the future of markets. It can help analyze information and improve productivity in decision-making, but it does not outperform indices. That said, if one day AI manages to anticipate the future, it will be the undisputed star of any Christmas meal. Source: cincodias.elpais.com