Artificial intelligence (AI) will not experience its watershed moment in equities trading, say market participants, as the industry seeks to avoid costly mistakes at the point of execution.
“There will not be a moment in time where there is a tipping point for adoption,” said Curtis Pfeiffer, chief business officer at Pragma Securities, in an email. “Improvements to execution algorithms are made gradually because the tools are relatively mature and been in use for 15 or more years. A new AI model in itself doesn’t guarantee better execution results.
“There is less tolerance for mistakes as execution algorithms handle real money. They are not recommending movies, like Netflix, or a new product to add to your shopping cart, like Amazon, where if the AI makes a bad recommendation, the cost of that mistake is negligible.”
According to an April report on equity execution by Greenwich Associates, less than 25% of traders currently utilise a system underpinned by a form of AI. Of those that do, 37% said that the algorithms led to an improved performance. The survey also found that buy-side traders have been shifting “modest to meaningful” amounts of trading business towards low-touch channels based around automation.
“When it comes to AI in trading, we are still in the very early innings, so it may be a few more years before we see widespread adoption,” wrote Richard Johnson, principal at Greenwich Associates and author of the report.
Pfeiffer added: “We are still in the early phases of AI adoption for execution algorithms and expect the use of new models and techniques made available for AI services to increase. The explosion of data, and the ability to analyse and derive meaningful insights in real-time – where to place the next order or when to reprice an existing order – is an area where AI tools can be of great value.”
Sylvain Thieullent, CEO of trading systems vendor Horizon Software, believes that AI applied to the execution block doesn’t really exist. “We believe that brokers can really take advantage of AI by capitalizing on an increased volume of data to provide enhanced client experience,” he said in an email. “More data and new AI-based strategies will offer improved execution with predictive indicators and insights from past execution scenarios.”
For Thieullent, the development of artificial intelligence solutions in equities has been down to both price and availability of data. “Today the industry is more mature and embraces the use of data which naturally leads the path to the first AI initiatives. But we shouldn’t get too excited about AI. Who is launching real AI projects? That’s an interesting question when many companies with valuations reaching billions of dollars stating they’re in AI when they’re just doing straight-through-processing and data management.”
An April survey conducted by financial data provider Refinitiv found that 90% of C-level executives are deploying artificial intelligence in pockets of their businesses, while around 75% said that machine learning had become a core component of their business strategy. Among the top applications for machine learning algorithms named in the study, 74% cited idea generation for trading and investment.