Abdur Amod, Head of Technology at Prescient Securities
The debate around AI and automation in asset management has largely focused on whether it will replace fundamental analysts, but that framing misses what may be the more consequential shift: AI could make fundamental research economically viable at a scale it has never reached before.
For decades, investors have faced a simple trade-off when picking companies, since they could either analyse a handful of them deeply or thousands of them superficially.
Generative AI is what begins to change that equation, because its greatest impact may not be in replacing fundamental analysts at all, but in extending their kind of deep research across far more of the market than a human team could ever reach.
Why this matters in South Africa
That shift could be particularly significant in a market like South Africa, where many small and medium-sized listed companies receive little research coverage, not because they lack potential, but because no one could ever justify the depth of work required to understand them properly.
Human attention is expensive, so it naturally concentrates where the money already is, like the banks, the miners and the other large companies that everyone follows. Markets tend to be most efficient precisely where that attention is deepest, so the room to understand a business better than the rest of the market widens wherever that attention thins out.
The greatest opportunity, then, may not lie in merely comprehending the JSE’s largest companies, but in the small- to medium-sized businesses that attract little institutional research, since no analyst could ever justify spending weeks on a company of that size, and it is precisely these under-covered names whose prices tend to drift furthest from reality.
There are, of course, limits, since less-followed companies can be harder to trade, and AI-generated outputs still require careful verification. Yet the old trade-off, of choosing between understanding companies deeply and covering many of them, is exactly what has started to erode.
Why depth lost ground
Fundamental analysis did not lose ground because it stopped working. It lost ground because it stopped scaling.
If understanding more companies meant hiring more analysts, then depth always had a ceiling created by the cost of hiring analysts. Quantitative investing did not have the same ceiling. Once a model was built, it could be applied across an entire market without hiring a new analyst for every company.
As the economics of investing evolved, routine analytical work became increasingly systemised. Signals that once required hours of manual effort could be captured and ranked by models at scale. Deep human coverage survived where the money justified it, namely the large, liquid, well-known companies. Proper fundamental coverage for smaller to medium-sized firms became too expensive to sustain.
Measuring is not understanding
What most people overlook is that quantitative investing never ignored words completely. For years, systems have processed news articles, regulatory filings, and earnings call transcripts. But “read” is the wrong word for what these systems were doing. They measured language rather than understanding it, counting positive and negative words, scoring sentiment, flagging risk-related phrases, and tracking changes in wording. A rich and contradictory earnings call could be reduced to a single sentiment score.
Experienced analysts do something different. They follow management’s argument, recognise when explanations change, identify inconsistencies between what executives say and what the financial statements reveal, and judge whether today’s promises align with yesterday’s commitments. Their advantage has never been reading documents; it has been understanding them.
What AI changes
Generative AI changes the equation because it works with meaning in a way older systems could not.
Give it an annual report and it can follow the argument, compare today’s commentary with last year’s, summarise the business, identify risks, and ask whether the explanation matches the numbers. That is much closer to what a fundamental analyst does than what a word-counting model does.
This is the real shift: AI does not make fundamental analysis irrelevant; it makes it scalable.
The work that once consumed an analyst’s time, reading, comparing, questioning, and connecting statements across documents, can now be done across a broader universe of companies. The traditional limitation of fundamental analysis was never a lack of insight, but a lack of reach.
For the first time, the choice between deep and wide is no longer as fixed as it used to be. The ability to analyse deeply and at scale is becoming increasingly achievable.
The honest limit
Older systems were limited in their capabilities, but they had one advantage: consistency. Ask a system to count negative words and it gives the same answer every time. Generative AI is different. It works with probabilities rather than certainties. This means it can give slightly different readings and, at times, be confidently wrong. In financial services, that distinction matters. A misread footnote or misunderstood disclosure is not a small problem when real money is at stake.
This is why AI should not be treated as an investment decision-maker. Its outputs still require human oversight, challenge, and context. Yet this does not weaken the case for AI; it strengthens it. AI makes fundamental analysis more relevant because it expands the field of companies that can be understood, while still leaving judgement where it belongs: with the analyst.
The future may not belong solely to those who can count the market fastest, but to those who can understand it widest. In a world filled with data, the biggest advantage may no longer be speed but rather scalable understanding.
ENDS
