AI and proxy voting: a Matrix moment for stewardship?

When The Matrix was released at the turn of the millennium, artificial intelligence was still largely confined to science fiction. Two and a half decades on, AI has become a routine part of daily life, from virtual assistants to traffic management, and its arrival in investment stewardship now feels less like fantasy and more like an inevitability.

That shift is now playing out in proxy voting. Earlier this month, JP Morgan Asset and Wealth Management announced it had cut ties with external proxy advisers in favour of an AI-powered in-house tool. The move raises a central question for stewardship: does greater automation strengthen investor oversight by improving scale and consistency, or does it risk introducing new forms of opacity?

Political pushback

JP Morgan's move did not come as a surprise to analysts who follow the manager. Jamie Dimon, chief executive of JP Morgan Chase, has previously criticised proxy advisers, warning that they exert undue influence over shareholder voting. The manager's new tool, Proxy IQ, is designed to handle volumes of data similar to those previously processed by proxy advisers, with JP Morgan retaining responsibility for voting decisions, which are executed by Broadridge.

But the announcement also comes at a sensitive time. The proxy voting industry has come under increasing scrutiny in the US, with president Donald Trump accusing firms of providing “politicised advice” and empowering the SEC to curtail their influence.

The theme was echoed in January by Brian Daly, director of the Division of Investment Management. Daly criticised the oligopolistic power of proxy advisers and suggested imposing further requirements for firms to justify their methodologies on ESG and DEI decisions. He also warned that asset managers working with proxy advisers on ESG issues could be in breach of their fiduciary duties.

Criticism of the proxy advice industry is not new. Sarah Wilson, chief executive and founder of proxy voting firm Minerva Analytics, notes that advisers were already under pressure during the 2008 financial crisis, when they were accused of conflicts of interest between the companies they monitored and potential clients.

However, Wilson argues that the current debate risks conflating two separate issues. On the one hand is the market dominance of proxy advisers, with ISS and Glass Lewis accounting for more than 90% of market share. On the other is the accusation that providers are offering “politicised” advice.

On market concentration, Wilson agrees that there is a strong case for scrutinising the ISS and Glass Lewis duopoly, and that customised in-house stewardship services could introduce greater choice. She says asset owners have been more proactive in exploring alternative proxy advisers, while some asset managers appear slower to introduce competitive tendering processes.

Her concerns lie elsewhere. Wilson warns that outsourcing fundamentally subjective decision-making to AI risks creating problems around transparency and accountability, particularly in politically contested areas such as net zero strategies, ESG and DEI. European JP Morgan clients, she notes, may hold very different views from their US counterparts.

While acknowledging that AI can play a valuable role in handling large volumes of data, Wilson cautions that tools such as Proxy IQ could become black boxes. “We encourage plurality of service providers and think it is healthy for there to be choice, but we must not confuse technology with good outcomes. Clients should be asking questions about data provenance and regulatory oversight,” she says.

“We are regulated, we have codes of conduct and high levels of transparency about what we do. AI models are opaque. If anybody is talking about robo-voting, they should be very worried about robo-voting using AI-based proxy advice,” she adds.

The AI opportunity

Others take a more optimistic view. Gregory Elders, director and head of North America at Canbury Insights, a sustainability consultancy that applies AI to stewardship tools, challenges the idea that these systems necessarily lead to black-box decision-making. “At Canbury, we are fully transparent about the information we use, and it allows us to sift through data much more efficiently. That frees up time to focus on the decisions that really matter,” he says.

Georgia Stewart, chief executive of fintech firm Tumelo echoes this sentiment: “Next-generation proxy technology will enable fund managers to combine their own expertise and views with market norms. With this level of customisation, fund managers will effectively become their own proxy advisers, bringing the function back in-house,” she predicts.

Stewart also disputes the idea that in-house, AI-driven proxy advisers will be insufficiently regulated, noting that fund managers are already regulated as investment advisers. However, she acknowledges that increased autonomy comes with responsibility. AI tools must be properly trained, their outputs reviewed, and their use aligned with principles of transparency, auditability and fiduciary duty.

Gustave Loriot-Boserup, chief executive and founder of Compass Insights, a responsible investment data platform for institutional investors, agrees. He argues that AI enables investment managers to operate at a scale that was previously impossible. By bringing resources in-house and taking greater ownership of the process, managers could ultimately improve stewardship outcomes over the long term.

While views on AI in stewardship remain divided, its adoption appears inevitable. Elders predicts that the 2026 AGM season will provide an early test case, with some investors adopting AI tools while others continue to rely on traditional proxy advisers. Although ISS and Glass Lewis are also incorporating AI into their services, he believes they may be underprepared for the scale of competition they could soon face.

Unlike in the Matrix, many stewardship practitioners do not consider the introduction of AI to result in a loss of agency, but believe it offers opportunity to sift through increasingly complex volumes of data. Having said that, human decision making remains crucial when it comes to ultimately casting the vote.