Case Study
AI-Assisted Wealth Management
AI-Assisted Wealth Management
Focus time Percentage
+80%
+80%
Agentic FinTech Investment Platform &
Co-pilot Architecture
Co-pilot Architecture
Architected proactive agentic workflows and user-trust mechanics for an enterprise investment platform, establishing scalable AI interaction patterns that drove user engagement and platform adoption.
Situation
• Problem: Passive side-sheet chatbots created "blank page syndrome" and high cognitive friction for advisors.
• Impact: Reduced platform engagement and missed high-value portfolio opportunities.
• Impact: Reduced platform engagement and missed high-value portfolio opportunities.
Action
• Pivot from reactive prompts to proactive, event-driven action cues.
• Calibrate system trust signals to define when to automate vs. escalate to human oversight.
• Calibrate system trust signals to define when to automate vs. escalate to human oversight.
Outcome
✔️ Accelerated Task Completion
Reduced advisor workflow friction and time-to-action by automating routine portfolio monitoring and generating 1-click execution cards.
Reduced advisor workflow friction and time-to-action by automating routine portfolio monitoring and generating 1-click execution cards.
✔️ Increased Platform Trust & Adoption
Replaced passive chatbot interfaces with contextual agentic cues, increasing proactive client outreach opportunities and scaling advisor capacity.
Replaced passive chatbot interfaces with contextual agentic cues, increasing proactive client outreach opportunities and scaling advisor capacity.
✔️ Established Design System AI Standard
Scaled reusable agentic UI patterns and evaluation frameworks (FigAgent) across product workstreams to accelerate future AI feature deployment.
Scaled reusable agentic UI patterns and evaluation frameworks (FigAgent) across product workstreams to accelerate future AI feature deployment.
Proactive agentic interaction that reduces effort and builds system-user trust while enabling scalability
Starter prompt was the issue
An unfocused AI Advisor prompt in the side-sheet created friction with users expressing 'blank page syndrome'. I used Figma Agent to suggest alternative patterns, then checked for feasibility with the engineers.
Human testing, AI reasoning
Combining user testing (with target users) to validate the problem with the starter prompt was a real blocker to adoption, we used Figma Agent to generate a more proactive pattern that proved more engaging and trustworthy, instigating human-advisor interactions that scaled.
We used a Figma Agent Skill (FigAgent-Reasons.md) to validate the output.
We used a Figma Agent Skill (FigAgent-Reasons.md) to validate the output.
When to go Human
Mapping rules and logic for how an agent knows when to escalate to a Human-In-The-Loop scenario, whether the decision is invoked from an agentic workflow cue or a human operator using a natural language query.
