The AI stewardship assistant inside engage insights isn't a general-purpose chatbot bolted onto a dashboard. It's a retrieval system grounded in two things your team already trusts: your own engagement record, and the official regulatory sources that govern it.
Ask a general AI model about your stewardship programme and it will happily invent an answer. That's exactly what a regulated team cannot use. The assistant we built takes the opposite approach: every response is grounded in source material — your logged engagements, voting records, and the primary regulatory texts (SRD II, the ISR label framework, the UK Stewardship Code, SFDR) — and every claim it surfaces is cited back to where it came from.
In practice, that turns hours of manual work into a sentence. "Summarise our engagement with the utilities sector this quarter" returns a board-ready narrative built from your actual activity logs, with links to each underlying record. "Which of our holdings still lack DNSH evidence for the ISR renewal?" returns a list, not an essay. "What did we commit to in our last dialogue with company X?" pulls the exact note, dated and attributed.
The assistant is deliberately narrow where it matters. It will not fabricate a vote, invent an engagement that didn't happen, or paraphrase a regulation into something it doesn't say. When it doesn't have the evidence, it tells you — because in a stewardship context, a confident wrong answer is worse than no answer.
Under the hood, it maps every engagement and vote to the frameworks that govern your funds, so a question phrased in plain English ("where are we exposed on board independence?") resolves against structured, framework-tagged data rather than a pile of free text. That mapping is what lets the same record answer a portfolio manager, a compliance officer, and an auditor without three separate exports.
The point isn't to replace the analyst's judgement — it's to give the analyst back the day they currently spend assembling evidence. The rationale, the sign-off, and the strategic call stay human. The retrieval, the cross-referencing, and the first draft don't have to be.
This is the first of several deep-dives into how the platform works. Next up: how we generate a full SFDR Article 8/9 evidence pack from the same underlying record.
Amine Gharby
Founder & CEO, engage insights. Former Glass Lewis Director. Writes about stewardship, regulation, and building better tools for asset managers.
Follow on LinkedIn →