The application of AI in automating cognitive processes is on the rise. The launch of J.P. Morgan’s new retail trading platform underscores this trend, and highlights what is rapidly becoming the new normal: that big firms are capable of being as nimble as their smaller counterparts when it comes to deploying new business models. Merrill Lynch is even experimenting with an AI stock-picking tool to help the firm identify value in small-cap stocks that human analysts might have missed. The question is, can AI live up to the hype and deliver excessive gains for investors?
Indeed, Wall Street has long been a forerunner in harvesting the powers of new technology – and AI is no exception. A recent study by McKinsey found that a full 28% of Financial Services companies have adopted AI technology, placing the industry third behind only High Tech & Communications and Automotive & Assembly. Prominent examples include BlackRock’s reliance on AI for identification and exploration of non-intuitive relationships between securities and market indicators, as well as for monitoring search engines for popular keyword inputs.
All of this is terrific news for technologists and financial advisors alike, as the fusion of these two worlds are likely to produce both technological innovation, and previously untapped alpha capture for years to come. However, the early adopters of AI need to understand not just where AI can boost innovation, insight, and decision making, but also where AI can’t provide value. Below is a simplified view of what AI is good – and not so good – at accomplishing:
What AI Can Do:
- Specific cognitive tasks
- Correct human bias
- Continuously analyze information
- Continuously monitor and gather public information
- Detect shifting sentiment within a specific domain of expertise
- Provide actionable, contextual and predictive insights
- Reveal hidden trends, bottlenecks or opportunities
- Help access relevant information faster than competitors
- Help manage idiosyncratic risk and opportunity caused by the digital revolution
- Generate alpha
What AI Can’t Do:
- Make decisions for you
- Understand general concepts
- Tell you what’s important to pay attention to without understanding the specific problem at hand
- Magically print money
The key takeaway is that AI is effective when used to scale human expertise in specific contexts. To wit, most AI models are trained through “Supervised Learning,” which means humans must label and categorize the raw data that gets fed into the system. However, this approach does not scale human experience, knowledge and effort. AI systems that scale human expertise must be seeded with a specific basis of understanding from which they can reason and learn at scale.This approach requires proprietary learning and natural language processing technologies, human expert networks and a deep understanding of the real world problem at hand.
Ultimately, AI is still very limited in its 'intelligence,’ hence increased adoption of the technology will continue to rely on human capital; that is to say, how firms build and scale their knowledge base. This will likely be a gradual shift, as it takes time to embed a culture that embraces the type of innovation and operational disruption that an effective AI delivers. But as the aforementioned examples illustrate, the gears are already turning.
Indeed, portfolio managers are currently experiencing a bit of a “golden age” when it comes to AI innovation – since the underlying technology is improving so rapidly, and consequently the outcomes (better investment decisions) are yielding better and better results. It bears repeating, however, that AI will not replace humans entirely. The advancement will simply serve to enhance specific human cognitive functions, not supplant them altogether.
For more on how AI and unstructured data are reshaping the investment landscape, check out Nasdaq’s interview with Accrete Founder and CEO, Prashant Bhuyan