Seven years ago, the massive money manager AllianceBernstein set out on a mission to transform itself with AI and data science.
"At the time, I went to my CEO and said, 'Look, we really need to rebuild our data science capabilities because our industry is changing and for us to be competitive in our industry," Andrew Chin, who at the time was serving as AB's chief risk officer and head of quantitative research, told Insider. The changes would be far-reaching, from how AB collects data, to how the organization is structured, to how investment decisions are made, he said.
Over the next several years, Chin, now AB's head of investment solutions and data science, would build out an internal task force dedicated to machine learning and AI. The move would save AB hundreds of thousands of dollars and help its investors seek an edge with the $704 billion assets they manage.
Wall Street firms' AI efforts have been revitalized this year, thanks to the mainstream release of ChatGPT and other forms of generative AI. Other asset managers are turning a corner with AI and data science. BlackRock has also been experimenting with AI to help employees do their jobs better and faster, with CEO Larry Fink predicting it could boost productivity by 30%. Vanguard and Fidelity are also building out AI capabilities.
Chin lifted the hood on the asset manager's AI strategy and use cases. The applications include using AI to signal differences in portfolio companies' regulatory filings, which can indicate poor performance, and using algorithms to uncover changes in strategy shifts and manager terminations before news outlets.
Using AI signals to make investment decisions and time buys and sells
AllianceBernstein uses AI to sift through more than 400 company reports and filings daily to uncover potential risks in relevant companies, whether in AB's clients' portfolios or a potential direct investment. Underscoring this is natural language processing, a form of AI that can process human language in a way computers can understand. The NLP is used to create investment signals, which are alerts that could uncover poor performance or predict positive movement.
For instance, one signal compares regulatory filings for differences in company strategy or management. The reason is stocks underperform when there are a lot of differences because it means that the company has a lot of dramatic changes. Usually, that means there are challenges that they're facing," Chin said.
In one recent case, the NLP found significant changes among 10K filings of a retailer AB owned in one of its equity portfolios. Specifically, the retailer had focused on growing store count as a core part of its growth strategy, but AB's AI tool found that was no longer the case in the retailer's 2023 10K. Chin's team alerted the analyst covering the retailer. They determined that strategy shift would be negatively viewed by the markets in the short term, but the fundamentals were still strong in the long term.
Because AB is a long-term investment firm, the asset manager held its position — despite the stock price significantly underperforming. About a month later, the stock retraced some of the losses, Chin noted.
AI is increasing productivity while improving risk management
In addition to comparing company filings, AB uses natural language processing to monitor reports, news, and other written documents for insights in language to measure potential company performance. For instance, the level of complexity, or how difficult language is to understand, can indicate poor future performance of a company, Chin said. He added that analysts and portfolio managers use these signals for idea generation and to time buys and sells.
AB also uses NLP to summarize new offering memorandums, legal documents issued to potential investors in a private placement deal. The documents are typically 300 pages long, making it difficult for analysts to assess these new opportunities daily. With NLP, analysts can quickly assess twice the number of summaries for client portfolios in less time, Chin said.
It's not just front-office investors and analysts who are benefiting from AI. AllianceBernstein is leveraging AI and natural language processing in its back office, Chin said.
In one case, the firm's operations analysts determine whether AB can invest in certain bonds or stocks for retirement accounts it manages based on whether they comply with the Employee Retirement Income Security Act (ERISA).
The documents are about 150 pages each, and there can be as many as 50 new securities issues every day, Chin said. The firm uses NLP to parse the documents, recommend whether AB can invest in the security, and provide a way to trace back to the supporting text. The tool presents a suggestion to the operations analyst who makes the final call. The AI application allowed the teams to be 50% to 75% more efficient while improving risk management because the decisions could be validated against the supporting text, Chin said.
