Section 4: What investors need

When made available, new data gets used

Since 2021, the ground-breaking ESEF digital reporting format has been required under the FCA’s rules for UK main-market listed companies. After two years of ESEF reporting, the FRC published the results of a related survey of 160 investment professionals from both the buy-side and sell-side investment communities.

With the FCA’s requirement  still relatively new, it was no surprise that the FRC study showed that 78% of those surveyed were still using PDFs as a source of data. Investors’ demand for the data that is absent or difficult to locate  in PDF reports was still being met by commercial data aggregation services.

But even at that early stage, things were changing. The survey indicated that 36% of investment professionals had added an ability to analyse XBRL tagging and were already using it. The level of use was nearly 50% in the sell-side investment sector that research companies use to make buy/sell/hold recommendations. Accordingly, the investment community was signalling that when new digitally formatted information is made available, it is valued and gets used.

The subsequent flow of newly released AI tools is only magnifying this trend, and the demand for trusted digital data sources increases with every AI application that emerges.

But investment professionals are also discreet about the specific ways they are using AI. The investment world is a competitive arena where market participants are looking to deliver ‘alpha’, the investment term for outperforming the market. In this world, it’s natural for investment professionals to be guarded on precisely how they model data to support decisions. It would be out of character for them to make their own models available and reveal how they are using AI to analyse data.

The rise of the bots

However, it is clear that investors’ AI tools seek data from more trusted sources. The academic study referenced in the previous section delivered a startling finding relating to the digital-first annual reports within their sample. In total there were 175 separate identifiable AI bots interacting with the data and between them they generated 4.8 million automated requests.

Unsurprisingly, the busiest bot was ChatGPT which generated over 30% of traffic. Bing, Amazon, Baidu and Google ranked second to fifth. We conclude that 175 is such a large number of separate AI bots accessing reporting that it seems certain that many of these were created and used by investment professionals using Agentic AI, automatically extracting and placing information into models before human eyes engage on the information.

Agentic AI is different to LLMs and other forms of Generative AI. Agentic bots can be programmed to execute multiple tasks towards a goal with some degree of autonomy, such as populating a financial investment model. Incidentally, the FRC has conferred with large audit firms to produce a useful definition and comparison of LLMs, Generative and Agentic AI in the context of corporate reporting and audits.

Pulling data into models is just one way AI is being used. The emergence of Natural Language Processing has taken AI-use into the realm of psychology where new documents are scanned daily to detect signals relating to sentiment, risk and changes of tone to help inform sell-side recommendations.

So the battle to deliver ‘alpha’ from information drawn from corporate reports is keenly fought on several fronts and AI plays more than one role. For a broad overview of how AI, corporate reporting and investors interact this article is well written and thoroughly researched.

Investment professionals need the highest quality data they can get, so next we consider the gold-standard for investor-grade structured corporate reporting, XBRL.