The meteoric rise of artificial intelligence (AI) in the public conscience has caused many people to question what an AI-dominated future looks like. Will AI transform industries? If so, will it democratize or consolidate them? Will it create better or worse outcomes? Outlines of answers can be found in the world of finance which has been transformed in the last decade by the same forces driving AI: the diffusion of ever more powerful computing and the profusion of data. The experience of finance is both encouraging and sobering for an AI-dominated future. It suggests that AI will transform some (but not all) industries, that it will benefit larger players most, and that just as it makes individual players smarter, it may make the world dumber.

The world of finance is an obvious laboratory for exploring the potential effects of AI because information processing is the central function of financial markets. Unsurprisingly, financial institutions of all types invest heavily in technology and data well ahead of other industries in order to compete most effectively. Of course, the experience of finance my not fully illuminate the scope of newer large language models that have so impressed the world in the last six months. But the changing competitive dynamics within finance over the last decade provide clues about what will happen across many industries when AI becomes cheaper and more widely available. And regardless of how these newer versions of artificial intelligence play out, finance will always to be the canary in the coal mine for the rest of the economy.

First, it appears clear that AI can disrupt industry dynamics very quickly. Consider the asset management industry. Over the last 15 years, we have witnessed two significant disruptions that can be traced to the growing dominance of technology and data. First, the mutual fund industry has seen the rise of passive fund managers (i.e., managers who invest in indices with no analysis) and the decline of active fund managers (i.e., stock pickers). This shift has occurred remarkably quickly as data and technology made passive investing more competitive and made it more difficult for active managers to attain informational edges. In the last eight years alone, the ratio of passively-managed assets to actively managed assets has risen from 0.6 to 1.2 — a dramatic shift in market share. The ability of active fund managers to extract large fees (upwards of one percentage point of assets under management) has been clobbered as passive fund managers demonstrated their ability to approximate many active fund management strategies at one-tenth of the cost.

Second, the hedge fund industry has been transformed by the growing dominance of quantitative investing over traditional, fundamentals-driven long-short strategies. The ability to analyze large amounts of data quickly and create relatively short-term strategies appears to be beating the slower and deeper analysis that traditionally led to long and short investment decisions. These trends in finance suggests that an AI-dominated future can create outsized winners and losers in very short order.

At the same time, the experience of the financial world suggests that not everything changes as quickly as people predict. While the high-frequency world of financial trading with its confluence of macroeconomic, sentiment, and company-specific information has changed rapidly, the lower frequency worlds of wealth management and lending have changed considerably less.

The much anticipated ability of robo-advisors to eclipse the massive financial advisory complex has appeared to stall and may be reversing. It appears that the client side of finance retains a preference for humans. Lending, similarly, has not been transformed by AI nearly as much as was predicted and AI-powered lenders have faced considerable problems. The incremental amount of additional data to be processed on individuals and business credit may just not be as large or as useful as in financial markets broadly.

The power of AI to disrupt industry dynamics appears to be tightly connected to the nature of the information problems being solved. Financial markets are a multi-dimensional information problem that requires massive amount of data and computing power. Fields with similar properties, like drug design, may be ripe for AI disruption. But many fields, including those in the services sector and manufacturing, simply may not have the same relevance for AI — they may be more like wealth management or lending. The experience of the finance industry suggests that human-facing services where data is not abundant and fast-changing can remain largely intact in a world of AI. To be clear, AI can still have a large impact by improving decision making but it is more likely to be incremental (as it has been in wealth management and lending) rather than transformational (as it has been in money management).

The world of finance can also help us understand if AI will be democratizing or consolidating. Here, it appears that the answer is less equivocal. Where AI has been pivotal (i.e., in financial markets), scale and speed appear to be the critical determinants of success. When technology and data come to dominate, winners keep winning and the ability to invest in technology and data is the key differentiator. A smaller quant fund has significant challenges in acquiring data feeds and computing power relative to established players. Similarly, fees for passive investing just continue to decrease as larger players share the benefits of scale with investors thereby boxing out upstarts. For sectors of the economy where AI is transformational, scale can be expected to be determinative and hopes for a great unleashing of smaller players that challenge established players appear to be overstated.

What can the finance industry’s experience tell us about whether AI is good for humans? Here, the experience of the world of finance is more sobering. The displacement of active managers who were charging large amounts for little excess performance seems like a positive development that is worth cheering. At the same time, it does not appear that financial markets are doing their central task — the processing of information — much better and it could be getting worse. The rise of investors that either willfully ignore information (passive investors) or obsess about fast-changing information (quant funds) means that the hard work of processing slow-moving, ambiguous, firm-specific information may be getting neglected. As data and computing come to dominate, industries may come to rely excessively on hard data that is fast-changing (e.g., stock price movements, real time credit card data on spending). Meanwhile, softer data (e.g., the future prospects of firms, the quality of management, the longer run consequences of pricing strategies) can be subordinated and diminished — even if it is what really matters for markets.

I fear this last lesson may generalize particularly well. The ability to analyze hard data in unstructured ways that are not directed by humans — the hallmark of AI — promises to transform the world in many ways, just as financial markets have been. But that transformation may be limited to settings where data is abundant and fast-changing. Moreover, the winners will be the largest firms able to invest in the computing power and data to create differentiated strategies. And the premium on the ability to consider softer data could fall in the short run even if, ultimately, it is what matters the most.

Can financial markets figure out how to capitalize on the wonders of AI and not neglect these more fundamental issues? The current equilibrium appears to be a financial market dominated by large players providing commodity services relatively cheaply but that neglects the processing of softer information. The challenge for the world of finance — and perhaps all of us — is to remember that the hardest questions facing managers and leaders are not entirely determined by hard data. What will allow my enterprise to succeed in 10 years? How can I deploy capital most effectively so that we innovate to create products and services that can serve our customers better? Hard data will inform these decisions but it is unlikely to be entirely dispositive. These decisions require acts of imagination and conviction. Just as the ability to use hard data cheapens and becomes more efficient via AI, it is these acts of judgment that will rise in importance. To acknowledge the primacy of these human questions does not diminish how much AI can help us — it simply reasserts that AI is merely a technology and that the greatest rewards for managers and investors rests in these fundamentally human endeavors.

LeackStat 2023