Beyond the Buzz: Industry Experts Unpack AI’s Rise in Retail Trading
Artificial intelligence (AI) is rapidly transforming the financial markets, changing how trading works. The phenomenon is bringing new ways to analyze and predict market trends, creating a whole novel ecosystem. In this article, industry experts and chief executives discuss these changes, blending caution with enthusiasm as they delve into emerging opportunities.
In the first part of this review, we discussed how AI is used in retail trading, highlighting relevant trends, benefits, risks, and mitigation techniques. Today, we’re featuring insights from industry professionals on their projections for the future development of this trend.
Tom Higgins, CEO of Gold-i, drew attention to the crucial distinction between different types of AI used in trading.
He clarifies that while AI has become a widely used term, it’s essential to specify the type being referred to. He downplays the relevance of natural language-based AI, like ChatGPT, in the context of trading due to the limited natural language involved in trading activities.
“Well it really depends on what you mean by “AI”. Unfortunately, the term has become so abused and distorted over many years of incorrect use, that it is important to define which type of AI we are referring to,” said Gold-i CEO.
“The current buzz in AI has come from tools like ChatGPT, which is a type of AI that is trained with vast amounts of data from the Internet and is then coupled with natural language engines to understand the original question and to generate an answer in a language that us mere humans can understand. This type of AI is not very useful in the context of traders and brokers as there is little natural language to deal with and learning half the knowledge of the Internet is irrelevant,” he added.
Higgins recalled the importance of machine learning (ML) from structured datasets, particularly in risk management and business intelligence tools like the Gold-i Visual Edge platform. He explains how ML helps in identifying known and unknown patterns of trading abuse by analyzing input data. When known patterns fail to identify undesired outputs, ML-based AI steps in, recognizing new patterns that might cause losses.
“ AI (the ML type) comes in handy when you know there has been some undesired output (like excess broker losses), but none of the known “input” patterns were able to identify it. The AI system will analyse all the input data, when it sees this undesired output, and will find patterns that the broker did not even know they should look for. The “learning” bit (comes into play now as the AI system remembers this rogue input pattern and will now also search for this pattern with the existing fixed patterns. The system will continually add to, and adjust, the patterns it is searching for as it learns the input patterns that cause losses and those that do not.”
“The first stage of deployment of this type of AI is to alert the broker of these new patterns it has found, but not to take any action. The second stage is for it to take some action, like A-booking the client in question if they are B-booked, or even rejecting their orders if necessary.”
That said, Higgins acknowledges potential risks during the initial phase of AI implementation, such as false positives, where AI may identify undesired patterns inaccurately. However, he highlights that these errors tend to decrease over time as the AI analyzes more data and learns complex rules. He stresses the importance of having a clear audit trail to explain AI-driven decisions, cautioning against situations where decisions lack transparency, merely attributed to “Computer says no,” as this lack of explanation could lead to problematic outcomes.
“Initially there may be too many false-positives, where the AI identifies an undesired input pattern , but, in fact, this is perfectly OK. This will improve over time as it analyses more data and learns more complex rules,” he concluded.
David Pope, the founder of Speech Craft Analytics, provided a unique angle on how stock traders might employ AI in their investment decisions.
David highlighted how traders and investors are using AI specifically for Natural Language Processing (NLP) and Voice Analytics techniques, focusing on company earnings calls. Intersitinly, he believes that certain linguistic and vocal cues could provide insights into a company’s performance potential.
“Companies executives who use language with strong positive sentiment, low complexity, and greater use of numbers tend to outperform. When spoken with higher pitch and harmonics the message conveys confidence which also is predictive of future outperformance. Conversely, should management use a lot of filler words, express voice micro tremors, and slow their speech rate it conveys nervousness and trepidation. It’s likely that the speaker does not have the level of confidence suggested by the words alone.”
Kevin Hamilton, CEO and co-founder of Tiblio AI, also provided insights into the latest developments in AI applications for trading and the existing or evolving AI tools available to traders and brokers.
Firstly, Kevin distinguishes between deterministic AI, suited for automating precise actions in an investor’s account, and the potential integration of more complex decision trees fed by Generative AI (GenAI) or models like GPTs.
“A deterministic AI is best suited for automation of activity on an investor’s account which can include opening and closing positions. These things have to be correct and have to do what you would expect or want them to do. So these are more or less complex decision trees. Those decision trees, however, can be fed by GenAI or GPTs. For example, if a GPT discovers some information, for example, the investor can be informed and then approve a change to a portfolio. Which then would continue to be driven by the decision tree type AI mentioned above which will follow the investor’s strategy precisely.”
Hamilton then highlights existing AI tools available to traders, including trade copying tools and platforms.
“There are trade copying tools, and even more use of AI tools like Tiblio AI which have the ability to fully automate options trading strategies according to what the investor wants to do. Tiblio is also developing custom data sets to better inform investors based on their portfolios and watchlists.”
Regarding broker usage of AI, Hamilton notes that large institutions have historically used machine learning algorithms for statistical arbitrage in markets. He also mentions specific instances such as Crypto.com’s AI chatbot feature and Interactive Brokers’ “iBot” as examples of AI usage in customer support and trade execution.
“I’m not sure how much AI it uses, but it is a chatbot that has the ability to interpret your request and place trades in your account – at least for simple things like “Buy 100 AAPL at a limit price of $154”.”
When discussing risks associated with AI implementation in trading, Hamilton acknowledges the inherent risks. He likens this responsibility to the risks already present in trading platforms, like app crashes or poor trade execution.
Tiblio, Inc. is a fintech company that creates software that helps investors trade through their brokerage account.
Jan Szilagyi, CEO of Toggle, shared his perspective into how AI, specifically their integration of GPT models, is set to revolutionize Wall Street and the financial landscape.
“This is the AI moment in finance – it will transform our ability to respond to fast-changing markets. AI is going to completely transform Wall Street because it supercharges the model’s ability to understand user’s intent – enabling advisers to trigger available analytical tools or access long-forgotten but still relevant research in seconds.”
Szilagyi highlights AI’s potential to address the challenge of discoverability in vast datasets prevalent across financial institutions. He notes that significant human capital, including written research and valuable data, often remains fragmented and scattered throughout the institution, hidden away in document repositories accessible only through keyword searches.
Moreover, Szilagyi stresses the issue of assembling scattered data pieces into a coherent and useful form, drawing a parallel to the “IKEA of financial information.
“You don’t know what to search for when answering questions like “What happens when inflation peaks” – do you look at core CPI, CPI, core PCE, … GPT-4 coupled with an analytical tool like Toggle can solve this issue. We are training GPT Models to help banks effectively revitalize their existing store of research and data, and enable any user across the institution to access relevant data OR analysis in seconds.”
Looking ahead, Szilagyi anticipates that by coupling advanced AI models like GPT-4 with analytical tools like Toggle, these AI systems will evolve into full-fledged analysts. These AI-driven analysts won’t just retrieve facts but will also conduct customized analyses, providing advisors and traders with personalized insights and decision-making capabilities.
“That will be the next frontier beyond merely searching for existing data: enabling advisors to perform custom analysis and generating personalized insights on that basis. In the future, a financial advisor is able to quickly run some basic analysis on any portfolio ahead of (or during) a call with a client.”
Meanwhile, Jonas Schleypen, the Co-founder & CEO of Hoc-trade, sheds light on the profound impact of Artificial Intelligence (AI) in transforming the trading industry and its applications, particularly for retail traders, brokers, and institutions.
Schleypen clarifies that AI in trading doesn’t entail fully independent, self-learning algorithms taking over entire trading operations; this domain remains primarily dominated by institutions. However, AI applications offer an array of valuable trading assistants that were previously unimaginable for retail traders.
“While AI was mostly focused on investments previously, such as in the form of portfolio composition and research, it now also enters the active trading market for retail traders. Traders should not mistake the use of AI with fully independent, self-learning algorithms, taking over their whole trading – this is an area which will be dominated by institutions with deep pockets – but AI applications enable a whole range of highly valuable trading assistants previously unthinkable of for retail traders.”
One intriguing application highlighted is Behavioral AI. These smart algorithms can detect emotional and biased behaviors in real-time, drawing from extensive trade data. This benefits both traders and brokers.
“Being trained on millions of trades, the Hoc-trade AI for example can real-time detect such detrimental behaviors, warn traders or the dealing desk, and provide understandable explanations and suggestions. Traders and brokers alike can benefit from such applications. Brokers have a powerful instrument for trader retention, activation, and differentiation, while traders can understand the effect of their trading psychology on their trading and are directed to the most important improvement areas.”
Schleypen was also excited about AI’s potential in Copy Trading. AI-driven filters can help followers identify and avoid unprofitable behaviors displayed by signal providers, allowing users to capitalize on strengths while mitigating weaknesses when selecting traders to follow.
“Selecting the right traders to follow is still one of the biggest challenges today. Signal providers are oftentimes very skillful traders, however, may have misaligned risk appetites to their followers, show human emotional behaviors such as everyone else, and may have their own weaknesses in trading. With AI, followers will be able to filter such unprofitable behaviors. You may like 5 signal providers, however direct the AI to filter any emotional behaviors, do not copy them when they are revenge or over trading, or when they are fighting the trend. AI will give users the chance to only follow the strengths of other traders, but not their weaknesses.”
Moreover, brokers are increasingly integrating AI tools into their operations and offerings, finding direct benefits to their bottom line, he added.
“Brokers are able to respond to their users’ requirements in ways not possible pre-AI solutions. For example, the Hoc-trade AI TradeMedic solutions screens trader situations, and if needed by the user, automatically generates a ‘medical report’ for the trader, outlining the individual trading issues, their severity, and performance effect, supported by LLM models for better understandability. Thereby, brokers support and educate their clients, while the broker directly benefits from reduced churn rates through its outstanding customer service.”
Finally, social trading network eToro provided us with insightful data regarding their retail investors’ perceptions and usage of AI in the investing landscape.
In terms of investing in AI, eToro’s most recent Retail Investor Beat survey reveals that 26% of U.S. retail investors perceive digital transformation, including AI, as a long-term trend in the market. eToro’s U.S. Investment Analyst, Callie Cox, notes that AI-specific stocks have been gaining momentum, particularly as tech stocks faced pressure due to concerns about interest rates. Investors are strategically turning to AI-related stocks, recognizing them as potential value investments.
Regarding their use of AI technology, the survey indicates that one in five U.S. retail investors are already leveraging AI technology to assist in selecting or adjusting their investments. Additionally, 50% of U.S. investors are ready to embrace AI technology, primarily as they believe it represents the future of investing. Other top reasons for embracing AI in investment decisions include the desire to save time on research (42%), confidence in AI making superior decisions compared to their own judgment (37%), and the belief that AI can outperform traditional fund managers in selecting better investments (37%).
The survey findings also highlight a notable trend among younger investors. Nearly half (46%) of investors aged 18-34 are already using AI technology to either adjust or select their investments, up from 31% in the previous quarter. Furthermore, adoption rates among investors aged 34-44 and 45-54 are on the rise, with the percentage of investors using AI technology in these age groups increasing from 38% to 45% and 12% to 19%, respectively.