From Numbers to Narratives: Screening Using Generative AI

For the past few decades, fundamental investors have steadily professionalized the art of screening. What began with simple filters on earnings, revenues, margins, and balance sheet strength has evolved into more refined quantitative frameworks and such screens are now table stakes. Joel Greenblatt’s Magic Formula is a classic example of this maturation: a small set of intuitive metrics, systematically applied, capable of narrowing thousands of companies into a manageable shortlist.

More recently, factor investing, smart beta, and ever more sophisticated financial databases have pushed this process further. Today, we are extremely good at screening what can be counted.

Until recently, what could not be counted remained largely outside the screening process. Management philosophy, corporate culture, strategic posture, and the way executives talk about competition or customers were left to narrative judgment, intuition, and time-consuming deep reads. These qualitative dimensions mattered enormously, but they resisted systematization. Investors generally reviewed them after the screen, not through it.

Generative AI changes that ordering. For the first time, it has become feasible to translate qualitative ideas into structured signals that can be searched, compared, and filtered at scale. Language models can ingest earnings calls, investor presentations, annual letters, internal memos, and even job postings, and extract recurring themes, priorities, and patterns of thought.

Let’s take Blue Ocean Strategy as an example. At its core, it emphasizes creating uncontested market space, redefining the basis of competition, and shifting the value curve away from industry norms. These ideas can be decomposed. Does management consistently emphasize non-customers rather than share gains? Do they talk about eliminating or reducing features the industry takes for granted while raising or creating others? Is pricing framed around accessibility and adoption rather than premium positioning? Do acquisitions and product launches cluster around adjacency creation rather than consolidation? Each of these elements is qualitative, yet each leaves linguistic and behavioral traces that AI can detect and score.

You are no longer just screening for cheap or fast growing. You are screening for a way of thinking.

Run that logic across thousands of companies and the screen starts to look very different. Firms that repeatedly articulate customer pain points ignored by incumbents, that describe competition in abstract or dismissive terms, or that frame innovation around simplicity and reframing rather than incremental improvement begin to surface. At the same time, businesses that talk incessantly about peers, market share battles, and marginal efficiency gains get filtered out. You are no longer just screening for cheap or fast growing. You are screening for a way of thinking.

The same applies to rule-breaker or contrarian investing styles. Management teams that explicitly reject industry orthodoxy, that frame regulations or cycles as opportunities rather than constraints, or that show comfort with temporary unpopularity can be identified through their language and decision patterns. Capital allocation choices, tone around risk, and the consistency between stated philosophy and actual behavior can all be assessed systematically rather than anecdotally.

These tools won’t replace judgment but they dramatically expand the funnel. Instead of relying on chance encounters or heroic reading efforts to find philosophically aligned businesses, investors can now let qualitative screens do the first pass.

Looking ahead, the next generation of these capabilities will likely move beyond static screening. We will see dynamic, time-series analysis of qualitative change. How does a company’s strategic language evolve before margins inflect or returns deteriorate? When does cultural drift show up in internal communications before it appears in the numbers? How early can shifts in risk appetite, competitive mindset, or capital discipline be detected relative to market perception? Eventually, qualitative signals may be tracked with the same rigor and historical depth as financial ratios.

In that sense, generative AI does not make investing less human. It does the opposite. By systematizing the first layer of qualitative understanding, it frees investors to spend more time on synthesis, judgment, and conviction. We spent decades learning how to screen numbers. We are now beginning to screen ideas.