Building a 0-to-1 AI Platform for 5,000+ Financial Advisors
80%
Reduction in documentation time
0%
Reduction in documentation time
0%
User satisfaction score
0%
Advisor adoption in 6 months
0%
Improvement in data completeness
The Problem
Financial advisors at Northwestern Mutual spend significant time after every client meeting manually writing case notes, drafting follow-up emails, and entering data into planning systems. For advisors averaging 4-6 client meetings per day, this documentation burden consumed roughly 1.5 hours per meeting — time that could be spent with clients.
The company needed a way to automate this workflow without sacrificing accuracy or compliance in a highly regulated financial services environment.
My Approach
I owned this product from pilot through production launch as the PM on the firm's highest-priority AI initiative.
Defining the product vision: I started with advisor interviews and shadowing sessions to map the full documentation workflow. The insight that shaped everything: advisors didn't just need faster note-taking — they needed structured data that could flow directly into downstream planning applications to reduce the burden of manual notetaking completely.
Architecture decisions: The platform processes recorded Zoom meetings through an LLM pipeline that produces three outputs: meeting summaries, client follow-up emails, and structured data extraction. I worked with multiple engineering teams to design a review workflow that gives advisors confidence in AI-generated outputs before anything reaches the final planning applications used to build the financial plan.
Navigating compliance: In financial services, every piece of client communication is subject to regulatory review. I partnered with legal and compliance stakeholders to design a phased rollout strategy that satisfied risk management while maintaining development velocity.
Driving adoption: Rather than a big-bang launch, we rolled out to advisor cohorts with progressively lighter onboarding requirements. I used qualitative user research alongside quantitative usage data to iterate on prompt design, UX review flows, and confidence signals.
The Outcome
Within six months of launch, 73% of eligible advisors had adopted the platform. Meeting documentation time dropped by 80%, saving approximately 1.5 hours per advisor per meeting. User satisfaction held at 96% — critical for sustained adoption in a population that's skeptical of tools that "automate" their client relationships.
What I Learned
Building AI products in regulated environments requires a different playbook than typical consumer or SaaS product development. The biggest unlock wasn't the technology itself, it was designing the right human-in-the-loop review patterns that gave advisors agency over AI-generated outputs in an intuative fashion. Trust was everything.