More integrated artificial intelligence tools are helping to improve stewardship by giving schemes and pension trustees better oversight of changes in voting policy, according to a new briefing note from Tumelo.
The fintech firm, which specialises in stewardship and investor voting, says AI is helping transform stewardship from an administrative exercise into a faster, more evidence-based process.
While many investment teams already use AI to draft emails, summarise documents and prepare reports, Tumelo says the real opportunity comes when large language models (LLMs) are connected directly to stewardship systems.
Tumelo says this allows stewardship teams to carry out complex analysis that would previously have taken weeks. It says one example is testing the impact of changes to voting policies before they are introduced.
The firm says it has helped clients model how tightening a board diversity policy from requiring 30 per cent female representation to 40 per cent would have affected voting outcomes across an equity index – although it has not revealed what this specifically shows. It adds that this type of analysis, that would previously have required weeks of manual work, can now be completed in a matter of hours, enabling stewardship teams to base policy decisions on evidence rather than intuition.
It adds that for pension trustees, this technology could improve the quality of stewardship oversight while giving investment managers more time to focus on engagement and decision-making rather than administration.
However Tumelo is keen to stress that AI should not replace stewardship infrastructure, but be a tool to help enhance it and should support rather than replace human judgement. This infrastructure includes voting policies, investment-grade data, voting records and auditable governance systems. Tumelo says that if a central LLM is connected to these, it can then generate stewardship reports, client responses and voting rationales while linking these output back to the underlying source data.
Tumelo’s briefing paper argues that this approach could make stewardship reporting more consistent and transparent for asset owners, while also helping managers answer trustee questions more quickly and test potential policy changes before implementing them.
It cites policy backtesting as an example, allowing stewardship teams to compare how revised voting thresholds would have altered historical voting decisions across different sectors or markets.
It adds that stewardship professionals should remain responsible for reviewing outputs and making final decisions, while the underlying infrastructure provides a full audit trail showing which policy was applied, what data was used and where the information came from.