Helping investment professionals explore how companies are exposed to emerging themes, and translate those insights into rules-based investment strategies.
Investors want to act on long-term themes like AI or clean energy, but turning a broad idea into a structured strategy was slow and manual. Engineering had proven exposure could be calculated — in a demo with no real product experience.
An LLM-powered workflow that links company business segments to themes and calculates exposure — so users can browse MSCI themes, define their own, and test them against any set of companies.
Launched in May 2026, replacing manual universe construction and turning a technically validated capability into a product that supports multiple investment-strategy use cases.
From a technical demo to a workflow designed around how people actually work
An exposure score measures how much a company is connected to a theme. If 60% of a company's business comes from AI-related activities, it has 60% AI exposure.
Engineering proved thematic exposure could be calculated. But the demo was a linear run of inputs ending in “calculate,” with no hierarchy and no sense of where you were or where to go next.
With the PM, I used Lovable and Figma Make to rapidly prototype alternative structures, navigation concepts and task flows.
Neither exploration stood alone. I combined the strongest elements of both with established Index patterns and our design system, so the product felt cohesive across the MSCI ecosystem.
AI accelerated exploration — not decision-making.
Prototypes came fast. The design work was deciding which ideas served users, met business requirements and could scale.
One flow from theme to exposure, built on MSCI's design system.
Themes start with MSCI's published library, each broken into sub-themes. My Themes sits right beside it, with a clear status (Live or In Review by MSCI) for every custom theme.
Start from an existing theme or a blank one. Name it, describe it, then scope each sub-theme with what's in and out of bounds, creating the rules the LLM uses to evaluate companies.
Name an experiment and upload a company list from a template. A verification step confirms every company before the calculation runs, so users catch mistakes up front instead of after a long wait.
Calculations run in the background, and users are notified when results are ready. Each company gets an exposure score and its top sub-themes, and any experiment can be re-run or saved to My Themes as the thesis evolves.
Launched May 2026
Want the full story? Get in touch for more examples
(next project)
Loan Manager