Last week, I blogged about this new paper from Glass Lewis entitled “AI and the Fiduciary Test: A Guide for Institutional Investors in Evaluating AI Proxy Voting Solutions” and I received a number of reactions. One response is this one from Nicolaas Koster of Proxywise AI: “Interesting paper, though I’d push back on a structural point. Glass Lewis treats proxy voting as if every decision requires the same depth of expert judgment. In practice, most ballot items are routine and policy-driven: they can be decided deterministically from disclosed facts and a client’s stated policy (e.g., voting against over-boarded directors).
That’s not a weakness of AI: it’s where AI is strongest, and it’s where human analyst time is being spent least productively today.
Genuinely hard cases (contested elections, novel shareholder proposals, go-shops where the mechanics matter) are a small share of any proxy season and deserve real judgment. The right question isn’t ‘AI or expert?’ It’s ‘which ballot items need which, and what architecture routes them correctly?’ The paper doesn’t really engage with that triage problem, which is where I think the interesting work is.”
Meanwhile, Glass Lewis has released this interview with two of its experts – Kaveh Beigi, VP of Digital and Content Strategy, and Arik Brutian, SVP of Artificial Intelligence and Data – which outlines what investment-grade AI requires in proxy voting workflows as a companion piece to its paper…
