MSCI AI Design System — Design Leadership Case Study
AI Design System
Design Leadership Case Study · 2026

One AI experience across every MSCI product.

How I led the design team in building MSCI's AI Design System from the ground up — turning a dozen competing AI experiments into a single, ownable AI language, the way Copilot is unmistakably Microsoft.

Design Systems AI / Conversational UX Cross-Functional Leadership 0 → 1
01 / Project

A shared language for AI at MSCI.

The MSCI AI Design System is the single source of truth for how every AI experience — chatbots, copilots, agents, AI-assisted flows — looks, behaves, and feels across the company. One system, unique to MSCI, adopted by product, engineering, and design.

My Role
Design Systems Leader
Set the vision, led the design team, and drove cross-functional adoption end to end.
What it is
An AI Design System
Principles, patterns, components, and a conversational master template — built for AI, not retrofitted.
In production
AskDocument & AskTPM
Two flagship chatbots shipped on the system — now rolling out across all verticals.
02 / The problem

Everyone was building their own AI.

Product teams across MSCI had an urgent, escalated need: ship AI capabilities fast to win a competitive edge. So each one started building — and each built a different AI.

The race to market created a quiet liability. Divergent AI journeys, visual styles, and interaction patterns were multiplying across product lines — every team reinventing the same chatbot, none of them recognizably MSCI.

The fragmentation

Before Many teams, many AIs

Scattered. Inconsistent. Off-brand.

AI v1 AI v2 AI v3 AI v4 AI v5 AI v6 no standard
Duplicated effort Inconsistent UX No shared brand Slower, not faster
After One system, every team

Unified. Consistent. Ownable.

Private-i Total Plan Data Plat. Indexes Analytics ESG AI Design System
Designed once, adopted everywhere One MSCI AI voice Faster to market
03 / The vision

Stepping in with a point of view.

When the need escalated to leadership, I didn't wait for a mandate. I brought a vision: MSCI shouldn't have many AIs — it should have one, and it should be unmistakably ours.

The brief wasn't "design a chatbot." It was bigger than that. If AI was going to show up in Private-i, Total Plan, Data Platform and beyond, it needed to feel like the same intelligence — a single personality, a single set of rules, a single visual signature. The way Copilot is to Microsoft, this would be to MSCI.

I framed the problem for leadership and product partners, secured buy-in, and set the design team in motion. The goal: not a one-off UI, but an owned, reusable AI design language — opinionated enough to be recognizable, flexible enough to live inside very different products.

04 / Explore

We studied the best, then explored hard.

An industry benchmarking study anchored the work in evidence. It became a decision engine — every finding turned into a design hypothesis we could test, keep, or kill.

Benchmarked
AI visual language
Compared the AI visual language across SaaS apps — Copilot, Rovo, Gemini and others — to pin down what actually reads as “AI” on an interface. This guided MSCI's own AI iconography, color palette, and visual style.
Studied
Conversational UX
Prompt entry, suggestions, streaming responses, and follow-ups.
Studied
Trust & citations
How best-in-class tools show sources, confidence, and grounding.
Studied
Surfaces & entry points
Side panels, inline assists, and full-screen AI workspaces.

The question underneath every exploration was deceptively simple: where should AI live, and how should it behave? We ran the benchmarking findings through several rounds of design, pressure-testing four distinct directions with the team before the system took shape. Here's the trail — what we tried, and why each survived or got parked.

The directions we explored

Direction A Kept · default
“What if the assistant docks beside the work, always in reach?”
Kept context visible while users conversed. Became the system's default surface.
Direction B Kept · light
“What if AI appears inline, exactly where the task is?”
Ideal for quick, contextual assists. Kept as a lightweight companion to the panel.
Direction C Kept · deep
“What if heavy tasks get a dedicated AI workspace?”
Right for long reports and analysis. Kept as the panel's scale-up state.
Direction D Parked
“What if AI is a floating command bar over everything?”
Felt disconnected from the work and competed with product nav. Parked.

One question, one principle. The directions converged on a single answer — one behavior, many surfaces — which became a core rule of the system: panel by default, inline when light, full-screen when deep.

The full field

figma.com · AI Design System Explorations 2025
AI Design System Explorations

Founded on real use cases. The explorations were never a styling exercise — every direction was tested against real asks from business units and product teams across the enterprise. That grounding is what took us beyond a single AI icon to a scalable component library: one system any team can build on, from an in-app AI insight to a full conversational chatbot.

05 / The system

The MSCI AI Design System.

The explorations converged into a real, documented system — principles, foundations, and a component library purpose-built for AI experiences and ready for teams to build on.

Foundations & tokens
Color, type, spacing, motion, and an AI-specific token set — the atoms every pattern is built from.
Conversation components
Prompt entry, message bubbles, streaming responses, suggestions, and follow-up chips.
Trust & citations
Source attribution, confidence cues, and grounded-answer patterns that make AI verifiable.
Surfaces & entry points
Side panel, inline assist, modal, and full-screen workspace — one behavior, many contexts.
AI states
Loading, streaming, empty, partial, and error — the moments that make AI feel reliable.
Usage guidance
Do/don't rules, voice & tone, and adoption guidelines so teams build it right the first time.
figma.com · MSCI AI Design System 2026

A system, not a screen. Documented foundations and a reusable component library — the difference between a one-off design and an asset the whole company can build on.

06 / The master template

One template for every chatbot at MSCI.

The system's centerpiece: a Conversational AI master template that any product can adopt — the canonical blueprint for chatbot experiences company-wide.

01 — Ask

Natural-language entry.

A consistent prompt surface with suggested starting points, so users always know how to begin — and the experience feels the same in every product.

02 — Answer

Grounded, cited responses.

Streaming answers with sources and confidence built in — the trust patterns from the system applied consistently, every time.

03 — Scale

Panel to full report.

The same surface stretches from a quick side-panel reply to a full, structured report — one template, many intensities of work.

figma.com · Conversational AI — Master Template

The blueprint. Adopt the template, plug in your product's data, and ship a chatbot that's instantly consistent with every other AI experience at MSCI.

07 / In production

From system to shipped.

A design system only matters if teams build on it. Working cross-functionally with engineering, we proved it twice — AskDocument and AskTPM — and set the pattern for rollout across every vertical.

AskDocument — built on the AI Design System
AskTPM — built on the AI Design System
Two products, one AI. AskDocument and AskTPM look and behave like siblings because they're built from the same system — not coincidentally similar, but deliberately identical where it counts.

"When AskDocument and AskTPM feel like the same product, that's the system working. The win isn't a single chatbot — it's that the next ten teams won't have to start from zero."
08 / Impact

What changed.

Not just a chatbot shipped — a way for the whole company to build AI.

Many → 1
A dozen competing AI experiments became one shared system. Divergent journeys and visual styles consolidated into a single, ownable MSCI AI language.
2 → all
Proven on two products, built to scale to every vertical. AskDocument and AskTPM set the pattern; the system is rolling out across product lines.
0 → 1
An AI design discipline, created from the ground up. From an escalated need to a documented system with leadership buy-in and engineering adoption.

The deeper shift is cultural. Teams that were each solving AI alone now start from a shared foundation — faster to market, and recognizably MSCI when they get there. The design system turned AI from a per-team scramble into a company-level capability.

That's the role of design systems leadership: not to design every screen, but to make the right design the easy default — at scale, across products, for teams I'll never sit next to.

MSCI AI Design System · Design Leadership Case Study · 2026