By Liam Hynes, PhD – S&P Global Market Intelligence: Systematic investing has always been a story of expanding information sets. Prices, then fundamentals, then alternative data. At each stage, investors sought to gain an edge by finding new information, organising it, and transforming it into repeatable signals. Today, a meaningful step change is underway. Not a new chapter, but a material advancement of one already in progress.
Alpha Has Always Migrated to Beta. Now however, the Speed Has Changed.
Factor investing offers the clearest precedent. Value, momentum, quality, historical growth: each began as genuine alpha. Over time they became beta factors, characteristics companies are simply penalised for lacking rather than rewarded for possessing. Alternative data followed the same path. Satellite imagery, credit card transactions, web traffic: each quickly competed away. The alpha has existed in harder to find places, buried in earnings call transcripts, regulatory filings, and corporate disclosures.
“We have been working on unstructured data for well over a decade,” says Drew Bowers, Senior Quantitative Analyst of S&P Global Market Intelligence’s Quantitative Research & Solutions (QRS) team. “The signals were always there. What has changed is the quality of the tools available to extract them and the compute power to do it at a scale that genuinely moves the needle.”
That democratisation comes with a familiar consequence. Alpha turns to beta as adoption spreads. It always has. What is different this time is the speed. As these processes are operationalised at scale, that early alpha simply becomes market beta. The firms that win will be those early on the alpha to beta curve. But as with every technological revolution, the greatest beneficiaries will not necessarily be the first movers. They will be the ones who adopt the technology, learn it deeply, and embed it into how they work.
From Analog to Digital
Natural language processing tools existed for years, but they were the analog era of text analysis. The dominant approach, bag of words sentiment, counted positive words, subtracted negative ones, and divided by total words to produce a net positivity score: a method as blunt as it sounds. Where that approach reads a document like a tally sheet, a large language model reads it like an analyst. It moves through each sentence, self-identifies the topic being discussed, assesses how material that topic is to the company, and determines the polarity of the language around it. The shift is not incremental. It is the difference between analog and digital.
What is also changing is the ability to deploy these models deterministically. By ringfencing a large language model to carry out a specific task against auditable data, the outputs become consistent, repeatable, and governable. As proprietary investment management IP is embedded into agentic systems, firms can encode years of research insight directly into their processes, producing outputs that are not merely useful but institutionally defensible.
“The core insight, that language carries investable signal, is not new to us,” says James Olejniczak, Quantitative Operations Director on the QRS team. What is new is the step change in compute and model quality that lets us act on that insight systematically, across thousands of S&P Global machine-readable documents, in a way that wasn’t feasible even a few years ago. “Markets are slow to process information that requires reading and interpretation, that creates an advantage for those who cannot only read faster, but interpret at scale.”
“The academic work known as Lazy Prices (Cohen et al. 2020) illustrates the opportunity precisely,” says Henry Chiang, Quantitative Researcher on the QRS team. The paper systematically examined year-on-year changes in the risk sections of corporate filings. It did not assess whether a risk had been added or removed, only just that the language had changed at all. Even that proved to carry meaningful signal. Now firms have the tools to actually analyse what that change means, and to do so at scale. “You could say ‘Lazy Prices’ is starting to be not so lazy anymore, the framework was always right. Now we have the tools to take it much further.”
Discretionary Analysis at Scale
The most significant consequence is not analytical but economic. The bottleneck has always been implementation. A research paper that once required months of engineering effort can now be translated into a working workflow in days.
“The efficiency gains are significant,” says Henry Chiang. “You are building discretionary analysis at scale. That is a fundamentally different proposition to anything systematic investing has been able to offer before.”
Today, AI can identify whether a company added a new regulatory concern, removed a supply chain risk, or introduced litigation exposure. Qualitative observations become measurable features. Human interpretation becomes scalable.
The Private Markets Frontier
If the implications for public markets are significant, those for private markets are considerable. Private company analysis has long been the domain of fundamental investors, not by preference but by necessity. Sparse financials, fragmented filings, inconsistent disclosures: the data landscape was simply too difficult to navigate for systematic strategies. The edge belonged to those willing to do the deep, time-intensive work of qualitative assessment.
“Private markets have always been the last frontier for systematic approaches, not because the signals weren’t there, but because accessing them required a level of manual effort that didn’t scale,” says Ilja Hauerhof, Director, New Product Development – Private Markets on the QRS team. “What changes now is that you can deploy the same agentic frameworks across private company data that you would apply to public markets. The information was always there. The infrastructure to interpret it wasn’t.”
The deeper opportunity is not confined to any asset class. Any investment house that believes it has an edge, whether in identifying management quality, assessing operational risk, or reading competitive dynamics, has historically only been able to apply that edge to the companies its analysts could physically cover. These tools change that constraint. Proprietary investment judgment, the kind built over decades of sector expertise and pattern recognition, can now be encoded into agentic systems and deployed consistently across entire universes of companies, public or private.
“If any investment house can encode its IP into an agentic system and deploy it at scale, that edge becomes systematic,” says Ilja Hauerhof. “That is what investing becomes, regardless of where you are deploying capital.”
What Comes Next
Prices created technical analysis. Fundamentals created quantitative equity investing. Alternative data expanded the investable information set. AI represents not a departure from that progression, but its next step, one that quantitative research houses have been building toward for some time.
“Better tools raise the bar for what rigour looks like,” says Drew Bowers. “The firms that benefit most will be those that combine genuine investment insight with the infrastructure to act on it at scale. The technology is only as valuable as the process it serves.”
The firms that succeed will not be those that replace judgment with AI. They will be the ones that use AI to expand the range of information their investment processes can understand, and who have spent the last decade building the foundations to do exactly that.
