Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Chances are, artificial intelligence can already do at least some of your job. But how much? The debate about AI in the workplace focuses on whether the technology will augment or do away with humans. But intuition, judgment, a gift for reading people — these are some of the skills hardest to mimic.
Think, for instance, about equity research. Analysts at Bernstein took the potentially career-endangering risk of handing typical research tasks to a comprehensive suite of AI models — including ChatGPT, Grok, Gemini and Perplexity. The models turn out to be really quite good at synthesising information. When fed historical earnings call transcripts, for instance, they identified major investor concerns and were able to assess how management had provided answers.
What they lack, on the other hand, is depth, the ability to put events into historical context, and to judge company prospects. That leaves AI, for now, as co-pilot rather than competitor, capable of doing grunt work while humans turn the results into actionable ideas.
That fits with the spiel heard from corporate executives across multiple industries, who claim AI will increase productivity rather than obviate the need for people.
In equity research, it rings true to a point. Research departments ought to be able to cover more companies, and cover them better. Small and mid-cap companies suffer from a dearth of analyst attention, which became more pronounced following European rules forbidding brokers from bundling research and trading. That is thought to contribute to the widening discount of UK small and mid-cap companies to the large-cap FTSE 100.
Analysts should not get too comfortable. Many professional investors say they value sell-side research not for its recommendations and conclusions, but for its analysis and data — in other words, the kind of content that AI seems to handle pretty well. When it comes to what really generates investment “alpha”, investors already claim to rely on their own fine judgment.
For those who fear eventual replacement, it is worth mastering what the AI cannot do. In the research world, humans may still be better placed to interpret soft clues on how a company is doing, for instance, by drawing conclusions from the tone and body language of executives during face-to-face encounters, or even their willingness to take investor meetings.
Such information has its drawbacks. Chief among them is that it is not available to the wider pool of investors. Moreover, AI advances quickly: at some point, AI-powered avatars capable of detecting non-verbal cues may be the ones attending company meetings. But as long as CEOs are humans, it is a safe bet some analysts will be too.
