AI or financial advisor? An Argentine experiment tests who achieves better returns

The financial industry, like many other activities, is undergoing a profound transformation. Advances in artificial intelligence and other technologies applied to financial analysis and portfolio management, such as big data, algorithmic trading, robo-advisors, and more recently, AI agents, have been driving innovations and changes in practically the entire profession for years.{#p-1787785160303-3412}

However, the speed of this process has accelerated considerably. During 2026, for example, Anthropic began integrating Claude with financial platforms such as FactSet, S&P Capital IQ, MSCI, and LSEG, allowing these systems to access market information, balances, estimates, and other professional databases within the same workflow.{#p-1787785160303-82035}

The advancement expands the range of tasks that can be automated and increasingly blurs the boundary between the functions traditionally performed by a professional and those that can be assumed by a technological tool.{#p-1787785160303-62374}

In this context, a report from the Institute of Economics at UADE compared for 24 months a portfolio managed through an algorithmic system with a fund managed by specialists, both exposed to the Argentine market.{#p-1787785160303-37960}

The result was a technical tie, although the evolution of each strategy showed significant differences depending on the context.{#p-1787785160303-33589}

What were the results of the experiment? {#p-1787785160303-46747}

During the bull market of 2024, the algorithm achieved a return of 130.2% in dollars, compared to 111.8% for the fund, in addition to recording lower volatility and a better risk-return ratio. However, in the bear market of 2025, the situation reversed, and professional management lost 14.6%, compared to 22.1% for the model. At the end of the two years, they finished almost tied, with 79.4% for the algorithmic portfolio and 80.9% for the fund managed by specialists.{#p-1787785160303-91394}

This does not mean that the profession has lost value decisively, nor that artificial intelligence can replace a financial advisor today. However, it does show that technological advances are beginning to compete with professionals in an increasingly large part of the investment process and could force them to redefine what value they bring within portfolio management.{#p-1787785160303-9268}

Moreover, the demand for this type of tools has been growing. A global survey by the CFA Institute showed that 84% of institutional investors would consider investing in a fund that primarily uses artificial intelligence to select assets.{#p-1787785160303-67023}

In this regard, Ignacio Meggiolaro, a leading lawyer in financial law and partner at Martínez de Hoz & Rueda, stated, "artificial intelligence will not eliminate professional judgment; it will exponentially increase the value of those who know how to combine technology with judgment, responsibility, and client knowledge."{#p-1787785160303-98327}

Is AI already on par with a professional manager? {#p-1787785160303-68835}

The comparison from UADE shows that the answer depends on the context. During the bull market of 2024, the algorithm not only achieved a higher return but did so with lower volatility and a Sharpe ratio of 2.68 compared to 1.85 for the fund, meaning the algorithm achieved more return for each unit of risk taken.{#p-1787785160303-47488}

However, when the market changed regime in 2025, professional management ended the year with a smaller loss.{#p-1787785160303-46916}

For Julián Colombo, director of Bitso for South America, the difference was not simply explained by the bear market, but by a design problem. At three points in 2025, the system concentrated all or almost all of the portfolio in just one or two assets. The most costly episode occurred in October, when the algorithm gained 30.9%, but a equally weighted portfolio with the same ten ADRs would have advanced 67.6%.

Meggiolaro agrees that AI has an extraordinary advantage in processing information, detecting patterns, and executing strategies with speed and discipline, although it still needs professional judgment to interpret contexts, manage risks, and understand the specific needs of the client.

Even the study itself reminds us that algorithms are not free from human decisions, as someone defines which assets to incorporate, which variables to weigh, and under what rules to operate.

Thus, rather than a competition with a definitive winner, the evidence points towards a hybrid management, with technology gaining ground in execution and analysis while professionals maintain an advantage in scenarios that the model has never observed.

Automation begins to modify tasks and labor structures within the financial industry.

Which financial tasks are starting to come under pressure

This hybrid model, however, does not imply that the transformation will be neutral for employment. In this sense, Meggiolaro warns that tasks historically performed by analysts, traders, and advisors will be automated at a much faster pace, especially when the work involves gathering information, making calculations, updating models, or executing standardized operations.

In this way, technology is already beginning to show that capability. Anthropic's financial tools can work between Excel and PowerPoint, consult institutional databases, and complete processes ranging from research to updating models and presentations within the same workflow.

The change is also starting to appear in the labor structures of the financial system itself. Standard Chartered, one of the largest international banks with a strong presence in Asia, Africa, and the Middle East, announced that it plans to eliminate more than 7,000 positions by 2030 while accelerating the use of AI and automation, mainly in lower value-added corporate functions.

The case is relevant because it shows that technological transformation is no longer limited to support tools for investors or analysts, but is beginning to directly modify how large banks organize their teams and which tasks they consider replaceable.

This does not necessarily imply the disappearance of the financial advisor. For Meggiolaro, it changes what value is expected from them. As processing data and executing tasks costs less and less, the importance of interpretation, strategy design, risk anticipation, and client support increases.

In this scenario, he argues that "the advisor who uses AI will have a huge advantage over the one who does not."

How Should Future Advisors Prepare Themselves

Adaptation, then, is less about competing with technology and more about learning to complement it. In this regard, Colombo argues that an intelligent agent brings discipline, backtesting, massive information processing, replicability, and less emotional decision-making, but can still fail when faced with situations that were never present in its training data.
His recommendation is not to compete with the algorithm "in what the algorithm does best". For those just starting in the profession, the differential will lie in developing the judgment to identify regime changes, interpret particular opportunities, and learn to audit the tools they use.
The case of UADE is illustrative because the system did not execute its instructions poorly, but rather correctly followed a rule that allowed for extreme concentrations.
Meggiolaro presents a similar challenge. A solid education in finance will remain essential, but it must be complemented with data, algorithms, and artificial intelligence. The professional will not only have to learn to use these tools but also understand their limitations and know when to question their results.
In that scenario, skills such as critical thinking, creativity, communication, and client knowledge gain value precisely because they are harder to standardize. The threat, then, seems to focus less on the profession itself and more on a certain profile of professional.
Those who limit their contribution to replicable tasks will be more exposed, while those who combine financial knowledge with technology will be able to significantly expand their analytical capacity.
Regulation Will Also Need to Adapt

The incorporation of AI also adds a discussion about responsibility. In Argentina, there is still no comprehensive regulation specific to artificial intelligence applied to financial advising, but, according to Meggiolaro, this does not imply a regulatory void.
The obligations of suitability, diligence, loyalty, client knowledge, risk profile determination, and adequacy of recommendations remain in effect even if an algorithm participates in the decision.
Therefore, the use of these tools should not dilute the intermediary's responsibility. In front of the client, he argues, it is not enough to state that "the algorithm decided". The regulated agent remains responsible for selecting, using, and supervising the system.
In this sense, the CNV advanced during 2026 in modernizing the suitability regime, with mandatory update courses and greater monitoring mechanisms, although so far it has not specifically incorporated artificial intelligence into the public examination program.
For Meggiolaro, it would be reasonable to advance in that direction and add knowledge about biases, data quality, model errors, concentration, and human supervision.

The discussion ends at the same point that opened the UADE experiment. The evidence from the 24 months does not allow declaring either of the two models superior and instead suggests that the combination of algorithmic discipline and human adaptability may offer a better investment design. AI can take on more parts of the process, but the challenge for the advisor will be to demonstrate its value precisely where automating a decision falls short.