Client

Fidelity Investments

Role

Senior Product Designer

Timeline

4 months

Platform

Desktop Web App

Domain

Fintech / AI

Year

2024

The Challenge

Fidelity's financial advisors manage hundreds of client relationships simultaneously. Each client conversation requires hours of preparation — reviewing portfolio changes, drafting personalized communications, checking compliance requirements, and synthesizing market data into actionable insights. With the emergence of large language models, there was an opportunity to dramatically reduce this cognitive overhead.

But the core design challenge wasn't "how do we add AI?" — it was "how do we add AI that advisors actually trust?" Financial advice is high-stakes, heavily regulated, and deeply personal. A single hallucinated data point in a client email could trigger SEC violations, erode client trust, or result in unsuitable investment recommendations.

I led the end-to-end design of this AI workspace, drawing on my experience with LLM interface patterns at Modguard.ai, compliance workflows from Deutsche Bank, and my deep knowledge of Fidelity's advisor tools and scheduling systems.

73%

of advisors spend 2+ hours prepping for a single client meeting

Internal research, n=48

89%

said they'd use AI tools only if they could verify every claim before sending

User interviews, n=32

4.2x

more compliance violations in AI-drafted content vs. manually written

Modguard.ai benchmark study

61%

of advisor time on emails is spent on formatting, not substance

Workflow analysis, n=24

01

Trust Through Transparency

Every AI-generated element shows its confidence score, source data, and reasoning. Advisors can drill into why the AI made a recommendation before acting on it. This pattern directly from my Modguard.ai work, where we learned that hiding AI confidence erodes user trust faster than low confidence itself.

02

Human-in-the-Loop, Always

AI content is always presented as 'draft' with clear editing affordances. Nothing is auto-sent without explicit advisor review. The workflow is AI proposes → Advisor reviews → Compliance checks → Advisor sends. This isn't just good UX — it's a regulatory requirement under Reg BI.

03

Progressive AI Autonomy

New advisors start with maximum AI guardrails. As they build confidence and the system learns their preferences, they can gradually increase AI autonomy — similar to how Tesla's autopilot unlocks features based on driver behavior. This directly addresses Rachel's need for gradual trust-building.

04

Compliance as a Feature, Not a Barrier

Instead of bolting compliance on as a gate at the end, it's integrated throughout the workflow. Real-time regulatory checks during drafting prevent violations before they happen — informed by my work on Deutsche Bank's compliance eLearning platform.

Domain Expertise

This project sits at the intersection of three distinct career experiences — a combination that's exceptionally rare in product design. Most designers working on AI-powered financial tools have either AI/ML pattern expertise or financial domain knowledge. I have both, plus firsthand compliance workflow experience.

When I designed the hallucination flagging patterns in this workspace, I wasn't guessing what advisors need — I'd already built and tested those exact patterns with real users at Modguard.ai. When I designed the compliance integration, I understood the regulatory framework from building Deutsche Bank's compliance education platform. This is the difference between a designer who researches fintech and one who has shipped in it.