Major asset managers including Balyasny Asset Management and BlackRock are increasingly leveraging artificial intelligence to extract investment insights from their own internal data, as traditional alternative data sources lose their competitive edge, according to a report by Business insider.
For years, hedge funds differentiated themselves by analysing non-traditional datasets – ranging from aggregated credit card transactions to mobile location data and satellite imagery – to anticipate earnings trends and commodity price movements. As these datasets became more widely adopted though, their value as a source of alpha diminished.
The rapid advancement of large language models has accelerated this shift. With firms now able to process vast amounts of publicly available information using AI, the focus is moving inward—towards proprietary data that competitors cannot replicate.
Speaking at the Future Alpha conference in New York, Jacob Bowers, a quantitative research executive at BlackRock, highlighted AI’s ability to organise and analyse unstructured information. He noted that some of the most valuable datasets now reside within firms themselves, including internal research, historical reports, and communications between investment teams.
According to Bowers, data that was once considered cutting-edge has effectively become commoditised in the AI era. In response, BlackRock has been deploying AI tools to mine internal archives in search of new investment signals.
The importance of proprietary data is not new. A 2019 report by Opimas suggested that asset managers could eventually monetise their internal datasets, while Robert Frey, formerly of Renaissance Technologies, has previously pointed to the firm’s extensive historical data as a key source of its long-term success.
Recent advances in AI, however, have made it significantly easier to unlock this value. At Balyasny Asset Management, for example, analysts are required to input research and notes into centralised systems, creating large volumes of text that can be analysed for patterns and signals.
Andrew Gelfand, a quantitative investor at the firm, noted that while efforts to harness internal data are not new, modern AI tools have substantially improved the ability to extract meaningful insights.
The effectiveness of this approach depends heavily on data quality. Internal datasets often reflect the thinking and decision-making processes of experienced investors, making them particularly valuable for training AI models.
However, even the most data-rich firms must continually update their inputs to reflect changing market conditions. As Mike Daylamani of Engineers Gate observed at the conference, high-quality data remains essential for building robust models.
