Yesterday, the WSJ reported that JPM’s asset management group will immediately stop using proxy advisory firms and instead use an internal AI-powered platform to analyze data from the more than 3,000 shareholder meetings it votes at each year — plus provide recommendations to portfolio managers and manage votes. JPM believes it is the first large investment firm to stop using proxy advisors entirely. This is all the more evidence that, as Liz said yesterday on CompensationStandards.com, we’re moving into our “fragmentation era” for investor votes. I worry that this is a “be careful what you wish for” situation, in that the potentially increased flexibility likely comes with more work and potential surprises.

It also means that it’s more important than ever to draft for the “bots,” which has been a recurring topic at our annual conferences over the last few years. Weil’s Howard Dicker shared a prescient recommendation during the panel “Proxy Disclosures: 12 Things You’ve Overlooked” at our 2023 conferences:

Within seconds of you filing your proxy statement on EDGAR, some [. . .] constituencies are using artificial intelligence (“AI”) to, among other things, give a rating, make a recommendation, prepare a summary or report, and write news stories.  Maybe even also to vote their shares automatically or to prepare a “fight” letter.

The takeaway is that not only should public companies be considering their varied constituencies, but companies also should be considering the AI technology that some of these constituencies are using. For example, academics and some analysts and investors have long performed computerized “sentiment analysis” on corporate disclosure documents, and some public companies have tried to make their disclosures more “friendly” to these algorithms. At an increasing pace AI is expected to become more sophisticated and its use more prevalent by these constituencies. Companies should take notice of this development, for the proxy statement and all other company disclosures.

Liz expanded on what it means to draft for the bots during the panel “The Proxy Process: Avoiding Surprises – On Time, On Budget & On Value” at our 2025 conferences. She suggested:

– Be cautious about excessive repetition of positive terms and vague disclosures that do not align with business performance. An overly optimistic tone can undermine trust with human readers and will be flagged by newer AI tools.

– Monitor how your disclosure performs in AI models, as their output can have a real-world impact on market sentiment and risk. Don’t overuse terms that signal uncertainty – such as “might” or “possible” – which may cause AI models to predict weak performance.

– Meredith Ervine