The rise of AI-powered language models (LLMs), such as ChatGPT and Gemini, has transformed how consumers access information, including personal finance and investment guidance. While these tools are powerful, accessible, and increasingly accurate, relying solely on them for do-it-yourself (DIY) investment decisions can be risky.
Reason 1: AI models lack personalised investment advice
LLMs operate based on generalised training data and some guardrails. They can provide educational explanations, compare investment products, and simulate hypothetical strategies. However, they do not have access to your Personal Circumstances, such as:
- Financial situation
- Risk tolerance
- Investment goals.
An investment adviser or a robo adviser conducts a discovery process and tailors investment recommendations to your personal and financial situation. AI tools, by contrast, provide generic information that can mislead if taken as guidance.
Reason 2: AI models can amplify biases
While AI feels confident in its responses, it does not truly understand investor psychology. It may deliver plausible-sounding advice that is:
- Framed with little or no regard for behavioural biases or emotional bias.
- Amplified by your biases and preferences, based on your previous activity with the model.
This drawback may lead investors to fall for the classic investment traps of overestimating their abilities, ignoring diversification, or chasing performance.
So, it is important that humans involve in portfolio construction and control the related decision-making process. While doing so, AI models can be used as tools by investment experts to be more effective.
Reason 3: Contrarian views are not typically considered
AI LLMs tend to echo mainstream opinions rather than introduce contrarian ideas. LLM models are trained on vast text corpora dominated by widely shared narratives, so their default recommendations usually reinforce familiar patterns. As a result, a truly contrarian investment thesis, one that goes against consensus market wisdom, will rarely surface unless explicitly prompted.
So, if you are not aware of a specific contrarian view, you might miss out on prompting the LLM about that investing view.
Reason 4: Risk of hallucinations and misinformation
Even with the latest advancements, AI LLMs occasionally generate incorrect or fabricated information, a phenomenon known as hallucination. For example:
- Offering contradictory portfolio strategies
- Using fabricated methodologies while constructing portfolios
- Incorrectly deciphering macro trends, systemic risks, geopolitical risks, etc.
Inaccurate outputs can seriously damage your financial well-being. There is a risk that LLMs can generate a response or construct a portfolio that leads to poor investment outcomes, posing a significant risk to managing life savings.
In contrast, a professional investment manager uses verified information and time-tested methodologies to construct the model portfolios. Even if the professional investment manager effectively utilises AI tools for some of her work, she still verifies the output of these tools. Thereafter, the Investment Committee discusses the constructed portfolios in detail before greenlighting them.
Reason 5: AI models have data boundary problem
AI LLMs are trained on historical text up to a fixed cutoff point, so they are unaware of events that occurred after that window. These AI models are limited by the static nature of their training and by the challenge of accessing up-to-date data, unless they use complex Retrieval-Augmented Generation (RAG) systems. In other words, it cannot account for market shifts that happened after training.
Moreover, there may be situations in which a significant geopolitical event or a financial systemic dynamic has occurred or played out, requiring an active involvement by a professional investment manager to respond to adverse market movements. An AI model with a data boundary problem will not be able to assist in this effort.
Reason 6: AI models have predictive limitations
LLMs function by generating answers based on historical patterns and existing training data, meaning they fundamentally “don’t understand the future”.
Because AI models rely entirely on past patterns, they offer no true foresight. They excel at describing historical trends, but cannot conjure genuinely new signals or warn of unseen risks.
AI models often rely heavily on their existing parametric knowledge, that is, historical facts on which they were trained, at the expense of reasoning over new narratives. So, whenever a new narrative challenges their memorised information, it leads to structural failures in reasoning about evolving conditions.
Because significant crises, such as market crashes, are infrequent, an LLM may have a minimal number of such examples in its training data. It cannot flag a crash the way a human might sense an unsustainable asset bubble.
In short, LLMs are powerful for summarising existing information but are fundamentally limited when it comes to forecasting future market moves or assessing forward-looking risk.
Reason 7: Failure modes for AI LLMs in crisis interpretation and unprecedented events
AI LLMs often struggle to handle novel or nonlinear crises. Because they base their answers on learned patterns, they can only guess by analogy. During black swan events, such as unprecedented financial or geopolitical disruptions, these analogies can be incorrect or incomplete. In turbulent times, AI portfolio construction can be misleading or dangerously simplistic.
Reason 8: AI models have no professional duty
When a robo adviser provides you with advice, they are professionally and legally responsible for how they arrived at that advice. However, AI LLMs provide information without accountability.
A robo adviser is licensed and operates within a regulated framework that holds them accountable to fiduciary standards and ensures they provide advice aligned with the client’s best interests. However, AI LLMs are neither licensed nor regulated entities.
Reason 9: Complete delegation of tasks not possible with AI models
A robo adviser conducts continuous or periodic monitoring of the portfolio, rebalances, suggests a different portfolio if the client’s Personal Circumstances change, and brings to the client’s attention any relevant action items. Essentially, it manages the investments it makes on behalf of the client.
However, the utility of an AI LLM in these aspects is limited or non-existent.
Related information
Read the answers to the related questions:
- Who constructs your portfolios — humans, algorithms, or artificial intelligence?
- Should I use a robo adviser or invest by myself?
- What does it take to make DIY (do-it-yourself) investments rather than hiring a financial adviser?
Also refer to the related knowledge resource:
QuietGrowth has been publishing content in this blog or in other sections of the website. Contributors for this content may include the employees of QuietGrowth, or third-party firms, or third-party authors. Unless otherwise noted, such content does not necessarily represent the actual views or opinions of QuietGrowth or any of its employees, directors, or officers.
Any links provided in our website to other websites are for the purpose of convenience, or as required by any such other websites. Unless otherwise noted, this does not imply that QuietGrowth endorses, is affiliated, and/or promotes any information, or products or services of those websites. Please read the advice disclaimer section of the website too.