NDAThis case study covers sponsor work protected by a non-disclosure agreement, specifically the sponsor company’s identity, product name, and prototype features are anonymized or generalized.
Outcome
We delivered a high-fidelity prototype and design-principle documentation to the sponsor’s leadership, defining an AI maintenance agent that keeps customer journey maps accurate and trustworthy over time.
Overview
The sponsor operates a customer journey management platform teams use to understand their customers’ experience across a product. We were asked to explore how an AI maintenance agent could be implemented into it to keep journey maps accurate.
The Problem
The platform’s users (CX strategists, designers, and product teams) manage, update, and analyze data inside customer journey maps.
Without a system for flagging staleness, they rely on manual and potentially inaccurate review cadences, so a map’s value degrades as maintenance grows with the volume of content.
My Role
I worked as a UX Designer & Researcher, leading concept ideation and exploration and contributing to research synthesis, wireframing, high-fidelity prototyping, and usability testing.
I also managed communication between the team, sponsor leads, and platform users through update meetings, update emails, and interviews, and mentored the junior designers on Figma onboarding and studio deliverable expectations.
Researching
Understanding What Makes an AI Maintenance Agent Effective in Existing Platform
To conceptualize the agent, we explored four research focuses: the platform’s state, the agent’s required components, the industry, and user needs.
To cover each within a tight time frame, we split into two sub-teams across two sprints, reconvening after each to share findings.
Current State of Platform
I audited the platform as a user, exploring its features across functionality and placement to inform where the agent would sit and how it would integrate with related features.
Current State of Agent Components
The partner team surveyed the AI components a maintenance agent would require: types of AI, categories of monitoring, and user trust.
Industry & Interaction Patterns
We then ran competitive analysis across direct competitors and analogous AI-driven SaaS platforms to understand the state of the AI agent industry.
4 Direct Journey-management Competitors
I audited one of the four hands-on, focusing on three aspects of its AI: user interactions, reach within the platform, and user flow pain points.
3 Analogous AI-driven SaaS Platforms
The partner team audited products that use AI to analyze customer data for feedback and insights.
We then examined interaction patterns within knowledge-verification platforms to surface the structural moves every modern verification flow shares.
User Interviews
Of 6 interviews with current platform users, I ran 3 as interviewer or note-taker; together they surfaced 8 themes the agent had to answer to.
Ideating
From Individual Sketching to a Sponsor-Selected Direction
We developed the concept through three stages: sketching, proposing, and wireframing.
Sketching Ideas
Each designer ideated in a Crazy 8’s sprint before we walked through everyone’s ideas and combined the most supported features through group whiteboarding.
During the walkthroughs I pitched a direction built on our recurring theme of user inputs, and went on to lead a rules-based approach to the concept.
Proposing Concepts
We presented six concept proposals, including the rules-based direction I led, each paired with the research evidence behind it. The sponsor selected two complementary directions.
Dual Feature Wireframing
With two concepts approved, we split the team evenly to develop both concurrently.
Concept A · My team
I led the team sketching variations of Concept A, holding to the sponsor platform’s UI patterns while providing access to Concept B’s features. Those sketches opened refinement through partner-team and sponsor feedback, toward a more compressed and decision-focused UI.
Concept B · Partner team
The partner team constructed Concept B’s wireframes. As I didn’t contribute to its ideation or construction, images aren’t shown.
Eventual Merging of Wireframes into One System
Wireframing revealed heavy overlap, as both concepts drew on the same context and evidence, so we merged them into a single agentic system.
Because the wireframes are built on the sponsor’s platform UI, they aren’t shown here to comply with the NDA.
Verifying
Prototyping and Testing the Unified Concept
We took the combined concept into high-fidelity prototyping, usability testing, and iteration.
Hi-Fi Prototyping
We built an interactive Figma prototype expressing the unified concept end-to-end through user flows, applying the insights the research had surfaced.
Because the prototype was constructed directly on the sponsor’s platform surface, the screens themselves aren’t shown here.
Usability Testing
To avoid setting false expectations about features that might not ship, the sponsor asked us each to recruit a UX Design classmate familiar with journey mapping as a substitute tester.
- Agent access
- Information organization
- Flow simplification
Takeaways
Design Decision-Making Related to Platform
This project was my first time designing within an existing product context and design system. The first half was exploratory and general regarding takeaways, but the second half focused heavily on the platform’s own needs, making it natural to narrow down findings to apply to the final product later on.
Narrowing down scope was a challenge subjective from project to project that I had previously addressed early on when constructing the first set of guidelines for Frogslayer. I now understand that specific factors, like a project’s theme of expansion, hints at its expected window for scope narrowing.