Over the past six months, our team partnered with KeyBank for a 10-week sponsored capstone project focused on exploring the role of generative AI in mobile banking. While KeyBank currently uses MyKey Assistant to streamline basic customer support with conversational AI, our challenge was to envision the next phase of intelligent, user-centered banking. We set out to design in-line generative AI features that improve trust, personalization, and efficiency, without overwhelming users or compromising control.
PROBLEM
SOLUTION
WHERE KEYBANK STANDS TODAY
CONTEXT
LITERATURE REVIEW RESEARCH
OPPORTUNITY FOR DESIGN
USER INTERVIEWS
KeyBank’s existing app relied heavily on user-initiated actions, offering limited proactive or personalized support, resulting in a disjointed experience that didn’t effectively guide or assist users through key financial tasks.
Most AI features in banking apps are hidden behind chatbots or complex settings, leaving users without timely, personalized help when they need it most.
Designing in-line generative AI features that act as intelligent, invisible teammates.
By embedding helpful insights directly into users’ banking flow, our solution supports goal tracking, budgeting, and automation while prioritizing clarity, transparency, and user control.
MyKey Assistant, streamlining basic customer support with conversational AI.
MyKey currently allows KeyBank users to check balances, make payments, and get quick answers, but it depends entirely on users initiating interactions and doesn’t offer proactive or personalized support.
We conducted a literature review to understand why proactive support is important now, explore key applications in digital banking, and identify challenges to inform the redesign of KeyBank’s MyKey.
How might we utilize GENAI within the KeyBank mobile app to produce a design that enhances user experience, provides meaningful financial guidance, and aligns with client needs and expectations?
28 screener responses, 8 interviews, 4 key topics
For our interviews, we wanted to learn about users’ everyday banking habits, their frustrations with current mobile apps, and their expectations around AI-driven support. We also shared examples of potential inline AI features to gather their reactions and understand how helpful or unfamiliar these capabilities felt to them.
TIMELINE
January - June 2025
(6 months)
ROLE
User Researcher
UX Designer
TEAM SIZE
Hailey Hjort
Brynn Morrison
Bianca Stiles
Me!
TOOLS
Figma
Miro
Google Survey
Google Sheets
WHY GENERATIVE AI IN MOBILE BANKING NOW?
AI is transforming banking across customer service, fraud detection, compliance, and wealth management. Banks are moving from rule-based automation to more dynamic, learning-driven systems like GenAI. GenAI enables more natural, personalized experiences by generating insights, text, and summaries based on user behavior and data patterns.
COMPETITIVE ANALYSIS
DIRECT COMPETITORS
Use conversational AI assistants (MyKey, Erika, Fargo)
AI lives in separate chat interfaces
Most AI is rule-based, with limited personalization
Many tools require users to initiate interaction rather than receive proactive insights
FINTECH CHALLENGERS
Design is mobile-first, with cleaner UI and faster navigation
Focus on 24/7 support and simplified money management
AI tools like Ally Assist help with spending insights and transaction history
Most fintechs use rule-based or FAQ-style chat, GenAI still underutilized
USER INTERVIEW FINDINGS
3 main findings.
Participants primarily use mobile banking apps for routine tasks like balance checks and bill payments but are frustrated by difficult navigation, hidden features, and ineffective AI tools like chatbots that often fail to resolve their issues.
THEMES
4 key themes.
We narrowed our findings to four main themes: users primarily rely on mobile apps for routine tasks, struggle to locate less prominent features, experience frustration with inefficient navigation, and feel skeptical about AI due to past negative experiences with chatbots.
Trust, Transparency, & Control
Users want to understand how AI makes decisions and maintain the ability to approve, edit, or decline automated actions, especially for sensitive tasks like payments and transfers.
Personalization & Relevance
Participants value financial insights and suggestions that reflect their unique spending habits and goals. Generic or robotic responses were seen as unhelpful or impersonal.
Budgeting, Spending Insights & Goal Tracking
Users are looking for simple, visual tools that help them track progress, monitor overspending, and stay on top of personal savings goals.
Efficiency, Automation & Streamlined Tasks
There is strong interest in automating repetitive or variable tasks and reducing the number of steps needed to complete common banking actions.
LOW FIDELITY
Early concepts for iteration.
We explored ways to surface relevant features contextually, simplify navigation for core tasks like transfers and bill pay, and integrate inline AI support to assist users without requiring them to search or ask for help.
WIREFRAME FLOWS
4 main flows.
We designed four main wireframe flows: budgeting, inline transfer, inline transaction details, and a dynamic dashboard that all aimed to surface relevant information at the right time, reduce friction, and support users without requiring them to dig through the app.
Budgeting
Guides users through the process of reviewing and adjusting their monthly budget, AI-generated recommendations to adjust their budget based on historical spending patterns
In-line Analyze Transaction
Allows the user to get more information on a specific transaction, This feature uses AI to give insights as to how the transaction compares to past spending behavior, whether it’s a recurring charge, or if it’s unusually high for that category.
In-line Transfer
Users are looking for simple, visual tools that help them track progress, monitor overspending, and stay on top of personal savings goals.
Dynamic Dashboard
Dynamic dashboard surfaces real-time insights and shortcuts based on recent activity such as spending trends, progress toward goals, or upcoming bills. A Weekly Snapshot included which provides a quick-glance summary of key insights.
Interaction & Micro Level
USABILITY TESTING
5 participants, 4 main tasks revolving around adjusting budget, clicking AI prompt, analyzing transaction, and viewing dashboard.
After creating our flows in Figma, we conducted a usability test to evaluate how effectively the designs supported users in completing key tasks and identifying areas for improvement.
We also realized that many users were unfamiliar with inline AI, so we designed an onboarding flow to introduce and explain the new features in a clear and approachable way.
Core Behavioral
Introducing a new design for KeyBank: an AI-embedded experience that combats the traditional input-output model by proactively surfacing support during critical tasks like bill pay, transfers, or budgeting. By integrating intelligence throughout the app, we aimed to reduce friction, anticipate needs, and transform banking into a more seamless and personalized experience.
The dashboard is where all your current, upcoming, and past trips are kept along with current trip reminders.
A search bar that lets you discover recommendations and popular spots nearby, with the ability to save places and easily share them with friends.
A polling system that allows users' to create and vote for options within their trip
Friends can chat and discuss trip plans whether that's through individual or group chatting.
FINAL DESIGN
Budgeting
In-Line Transaction Analysis
Dynamic Dashboard
On Boarding
The On-the-Go
Age: 26
Occupation: Healthcare Professional
Goals:
Quickly check balances and recent transactions
Make fast payments or transfers while in transit
Minimize time spent navigating the app
Needs:
Needs personalization, shortcuts to repeat common tasks.
Friendly UI with clear transaction info (“What is this charge?”).
Progress nudges on savings goals, optional advice.
The Money Multitasker
Age: 42
Occupation: Working Mom
Goals:
Track spending and budgeting progress regularly
Make informed decisions about where money is going
Needs:
Needs reassurance about AI capabilities and boundaries (read-only vs action-enabled), permissions, audit trails, and robust privacy settings.
Clear visualizations of spending and budget trends
USER PERSONA
Let's meet our users!
NEXT STEPS
REFLECTION
What's next?!
Designing with Unfamiliar Technology Requires Clear Communication
At the start of the project, we didn’t fully understand what generative or inline AI could look like in a mobile banking context. Through extensive research including a literature review, expert input, and user interviews. We not only deepened our understanding, but also learned how important it is to clearly communicate the role of AI in the user experience. We realized that successful AI integration depends on building user trust and providing just-in-time explanations for unfamiliar features.
Mentorship Accelerated My Growth
Regular check-ins with our project sponsors provided consistent feedback, helping us stay aligned with our goals and iterate with intention. Their mentorship pushed us to think more critically, consider technical feasibility, and strengthen our design decisions week by week.
Usability Testing is a Lens, Not a Checkbox
Through usability testing, I learned how small frictions in flow or clarity could significantly impact trust and task completion, especially when introducing new interactions or AI assistance.
Design System Integration
Align designs with KeyBank's existing design system for consistency and scalability.
Develop Higher-Fidelity Prototypes
Refine wireframes into high-fidelity mockups based on feedback
Expand Testing with Larger, Diverse Users
Wider range of banking habits, financial literacy levels, accessibility needs.
The small arrow is hard to tap, I want to click the whole row.
Personalization Preferences Are Strong, but Subtle
During A/B testing, users consistently preferred toggle switches for personalization settings.
Toggles feel official. Checkboxes seem temporary, like I'd need to hit save.
Insight Timing & Frequency Should Be Flexible
While weekly summaries were appreciated by some, others preferred monthly insights due to less frequent app usage.
I don't check my account weekly, so monthly insights might work better.
Clarity & Context are Key
Users found the inline AI features helpful, but wanted more context about how they work and what impact they have leading us to introduce an onboarding flow to build understanding and trust from the start
Strengthen Feedback after Actions
Ai-assisted budgeting and transfers were well-received when users could review and apprive actions, especially through Face ID. However, participants needed more context and clarity.
Users See Value in In-Line AI Support
Users responded positively to in-line AI features, noting they were helpful, relevant, and encouraged better financial habits, especially when surfaced a tht eh right moment within their normal banking flow
Non-intrusive pop-up
Customizable AI preferences
Subtle Icons
In-line budgeting
Easy-access tabs
Routine Banking Tasks Dominate App Use
Participants primarily use mobile banking apps for checking balances, paying bills, transferring funds, and depositing checks. Users are frustrated by time-consuming or inconvenient tasks, such as locating less prominent features.
I used my bank app to check my balance, pay bills, transfer funds.
Usability & Navigation are Persistent Challenges
Users consistently struggled with navigation, citing that key tasks like transfers, bill pay, and card freezing were buried in menus and required too many steps to complete.
I would like there to be a shortcut of commonly used functions on my dashboard so I can just face ID one click go into that function and it remembers the last function that I did within in…
AI Features are Underutilized
Experiences with AI-powered tools-such as chatbots and automated budgeting-are largely negative or limited. Users find chatbots unhelpful, often unable to complete tasks or resolve issues without escalating to a human agent.
I've used the chatbot before. I think I used it to try to contact an agent, but it brought me to AI virtual assistant. I was truing to see if it could get support for an unauthorized payment and it kind of went in circles.
Difficulty Deciding
Difficulty when reaching a consensus on destinations and activities that satisfy everyone. It created stress from trying to accommodate everyone’s preferences, which can lead to delays in planning.
Reaction to AI Use Cases
Determining initial responses to potential generative AI-powered features, such as personalized financial tips, explaining transactions, or smart saving goals.
Current Behavior & Awareness
Exploring how users currently manage their banking tasks and their general awareness of AI in finance.
Difficulty Deciding
Difficulty when reaching a consensus on destinations and activities that satisfy everyone. It created stress from trying to accommodate everyone’s preferences, which can lead to delays in planning.
Too Many Platforms
Many interviewees often rely on multiple platforms at once such as iMessage, Instagram, and the Notes app for communication and planning, making it difficult to keep track of information.
Usage Patterns & Frequencies
Investigating how often users would interact with their banking app if generative AI was implemented.
Data and Security with AI
Understanding concerns around data privacy, consent, and trust when AI is involved in financial decisions.
Why Now
McKinsey estimates GenAI could add $2.6T–$4.4T in annual value globally.
Banks are increasingly investing in GenAI to remain competitive and meet rising customer expectations.
Key Applications Identitfied
Customer service automation (chatbots, virtual assistants)
Personalized budgeting and financial advice
Fraud detection and risk assessment
Document processing and internal efficiency
Challenges Highlighted
Lack of explainability and user trust
AI bias and misinformation
User data privacy concerns
Direct
Direct
Direct
Direct
Direct
Fintech
Fintech
Fintech
Fintech