This article first appeared on Diligent Market Intelligence's Activism newswire. To register for a demonstration and trial of the product, click here.
Nick Capuano, partner at the firm, discusses how AI has surfaced as an active participant in proxy contests. Your recent study notes that major asset managers are starting to rely on AI-enabled proxy voting engines instead of traditional proxy advisors. We continue to hear that proxy advisor influence is eroding. How do you expect this to develop over the course of the next few years?
It’s very interesting right now because there are two dynamics taking place. You have regulatory pressure on the proxy advisors themselves and at the same time the evolution of technology like AI with more institutions and large asset managers likely to come out with announcements like JP Morgan did, deploying their own AI tools to analyze proxy materials. Right now, a lot of that technology is being used for clearer-day type of situations and proxy advisors still play a very prominent role in contested situations. One unexpected takeaway from this study, however, is that you could make the case that proxy advisor influence is actually increasingly important. Our study found that more than 70% of the time AI recommended in favor of the issuer, it was when they claimed victory from ISS or Glass Lewis. Kekst’s study also found that AI is currently more likely to support an activist’s case for change than an incumbent board and management team. Why the pro-activist tilt?
If you analyze the nature of communications, fight letters, press releases, anything that an activist puts out today versus what a company is putting out, you'll see that the activists use sharper language and are far more pointed in their criticisms. They can cherry pick data and present that data in a compelling light to showcase underperformance. Those are exactly the arguments that these engines tend to prefer, compared to what companies produce which tend to focus on procedural matters such as “they went public without reaching out to us first” or “they built this position in a suspect way.” The engines just don't seem to care about those types of arguments. They're not as precious around the rules of engagement. Can you expand on some of the biases exhibited by AI when you examined the subset of contested shareholder votes from 2023 to 2025?
What struck me most is that the engines themselves, the major LLMs such as ChatGPT, Claude and Gemini all exhibited such different biases. Claude gives more credence to an activist track record. ChatGPT may be persuaded by an activist's arguments around board entrenchment, for example. Gemini is more persuaded around not disrupting operational execution that's underway. These differences make it even more challenging from a communications perspective because you don’t necessarily know which frontier model each asset manager may be relying upon. That means you need to understand what each of these engines individually are predisposed to think about your company and only then can you optimize communications most effectively. If activists have honed sharper communication that’s resonating with such LLMs, how can companies best manage their own content to match activists’ more direct approach?
If you're looking at this from a company perspective, the greatest opportunity is that 50% of all the sources we found cited in these were owned or controlled communication channels: press releases, company websites, SEC filings, most notably. Recognizing that AI and natural language processing tools are parsing your press release before humans have had a chance to even read the headline means you have to pay attention to word choice, sentence structure, and how you're organizing that content in a different way than ever before. You don’t want to leave it to the engine itself to distill the most important messages. You should be structuring content with more bullets that make clear from your vantage point key takeaways about your strategy, your capital allocation plans, your growth plans, etc. That’s the key. It’s not about winning every single argument. It’s about making the most impactful arguments as clearly and succinctly as possible. Beyond that, your volume of communications may need to increase because if you were to run this analysis from one day to the next, you might get a very different result. You need to be almost constantly churning out materials on your website and other channels that reinforce your most critical points. Where are companies most likely to trip up on managing all these algorithms?
I think the common trap right now is not treating these AI engines seriously enough. Issuers need to recognize how quickly adoption of this technology is moving. Many companies are still treating this as if it's a retail-only chatbot phenomenon. It’s not. Investors of all kinds – even the most traditional institutions - are using these engines as a tool to do research on your company at any given point in time and are increasingly integrating it into their workflows. Since the introduction of the universal proxy card (UPC), individual directors have been viewed as more exposed due to the pick-and-mix approach. How does AI take such regimes into account?
It’s well noted that UPC has made proxy fights more personal but our findings generally suggest that these engines don't seem interested in those personal attacks. They care more about the higher-order company level messages from the activists and the companies alike. Another outcome of UPC has been the rise in the number of first-time activists. I see the potential for this to lower these entry barriers even further with screening tools that activists can and are using to identify underperforming companies. It's never been easier for someone who's interested in a first-time activist type of campaign to find a target.
