AI ASSISTANT

Desktop | Mobile Application

Role

Lead UX Designer

Project

TIAA Assistant – an R&D-driven intelligent assistant for mobile banking

Project Team

UX Designer (self), 1 Product Manager, 1 Front-End Developer (self), 3 Back-end Engineers, 1 Senior Executive Stakeholder

Platforms

iOS, Android, and Web

Project Overview

TIAA Assistant

As the Lead UX Designer, I was tasked with designing a new R&D-driven feature for TIAA Bank: an AI-powered assistant integrated into their mobile banking app. The goal was to explore innovative ways to serve customer needs around fraud prevention, account servicing, and financial education. This feature aimed to reduce call center volume, increase chatbot adoption, and build trust through a more intuitive and human-centered experience.

Challenge

Key Issues Identified

Despite initial interest in intelligent assistants, early prototypes lacked engagement due to:

  • Confusing flows and generic responses.
  • High escalation rates to live agents, especially around fraud and security topics.
  • Inconsistent voice and tone, leading to reduced customer trust in digital interactions.

ROLE & RESPONSIBILITIES

Leadership Focus

I led the end-to-end product design strategy:

  • Directed the full UX lifecycle from discovery to launch.
  • Facilitated cross-functional collaboration with PMs, engineers, compliance, and customer support.
  • Defined success metrics and partnered with data analysts to measure assistant performance.

Discovery & Research

User Journey Map

Methods Used

To uncover key pain points and opportunities

  • Conducted stakeholder interviews with multiple stakeholders.
  • Mapped the existing customer service funnel to identify automation gaps.
  • Ran user interviews and created personas to understand different user mindsets

Design Strategy & Ideation

Approach

With insights in hand, I:

  • Developed a conversation taxonomy to standardize assistant responses and intents.
  • Created scenario-based flows for high-impact areas such as lost cards, account freezes, and ACH transfers.
  • Introduced trust-building microinteractions, including empathic language for fraud-related use cases.
  • Designed a modular component system for reuse across multiple assistant personas.

PROTOTYPING & TESTING

Execution

  • Built rapid prototypes in Figma and integrated them into an internal test platform.
  • Conducted 2 rounds of usability testing (moderated and unmoderated), targeting users who had previously escalated from chat to a live agent.
  • Introduced trust-building microinteractions, including empathic language for fraud-related use cases.
  • Iterated on tone, delay timing, and visual cues based on user reactions to trust and confidence levels.

Starting the Design

Understanding the Flow

Initiating the design

After completing the user research, I started sketching a few design concepts. I labeled a few key elements on the page to ensure I captured all interactions and functions.

When designing the chatbot, I kept the user in mind, aiming for it to be an option that enhances the user experience without distracting from other features. The design was sketched with reusability as a focus, ensuring it's simple for both mobile and desktop, a decision that directly impacts our users' interactions.

Wireframes

Once I had a firm grip on the concept based on the sketch, I started creating the initial wireframes, representing a low-fi user experience flow.

Low-fidelity Mobile Prototype

Adhering to the design process, I created interactive prototypes for mobile and desktop to ensure the application would function as expected.

View Prototype

Low-fidelity Desktop Prototype

I designed a low-fi prototype of the desktop concept, and I wanted the content to expand the desktop version of the feature to emphasize the active chatbot.

View Prototype

Refining the Design

Mobile Mockups

Following the usability study, a few participants stated that they had a hard time with background blending with the chatbot.

With the blending issue in mind, I decided to add an overlay over the distracting background content.

Before usability study

After usability study

Desktop Mockups

After conducting usability studies, I noticed a trend in which participants did not initially see the chatbot feature. I wanted to emphasize the chat feature, so I moved it from the bottom right of the screen to the top left.

Before usability study

After usability study

Mockups

High-fidelity Prototype

Please click the button to see the high fidility prototype of the mobile experiecnce.

View Prototype

Mockups

High-fidelity Prototype

Please click the button to see the high fidility prototype of the desktop experience.

RESULTS & IMPACT

Business & User Outcomes:

After launching the redesigned assistant as part of the TIAA app's R&D rollout:

  • Reduced call center escalations by 28% in 3 months.
  • Improved assistant task completion rates by 42%, particularly around card-related workflows.
  • Increased user trust scores in post-chat surveys by +18 NPS points.
  • Assistant interactions grew from 1,200 to over 3,500 per day post-launch.

REFLECTION

What I Learned

  • If given more time, I would’ve expanded our emotional sentiment tracking to further personalize experiences.
  • This project taught me the power of aligning AI intent design with emotional UX—especially in sensitive contexts like fraud and finance.
  • I’ve since developed a framework for assistant trust design that I now use as a baseline for all conversational products I work on.

Key Takeaways

Highlights

  • Strategic focus on cross-functional alignment and trust design.
  • Demonstrated ability to drive measurable outcomes and user adoption.
  • Balanced conversation design with system scalability for future assistant use cases across the organization.