The Challenge
The client’s internal support portal was underperforming, creating friction for employees and burdening support teams. Employees often abandoned official workflows, instead contacting support through backchannels. This behavior skewed analytics, slowed ticket resolution, and hurt the credibility of the support system itself.
Our initial ask was to design a virtual agent that could redirect these inquiries. However, early stakeholder conversations revealed that their assumptions about user behavior were only partially grounded in evidence. Stakeholders believed that users were lazy or resistant to the process, but this explanation felt too simplistic and risked masking deeper UX failures.
Approach
1. Reframing the Problem
I guided stakeholders through a discussion on problem framing, elaborating that what we often perceive as “laziness” has deep ties to human psychology and problem-solving. Drawing parallels to cognitive bias and design constraints, I made the case that we should first explore user motivations and barriers before committing to a solution.
To their credit, stakeholders gave us space to investigate. We shifted from solutioning to problem definition, aligning our team around a hypothesis that the issue wasn’t just user behavior.
2. Research & Discovery
We conducted mixed-methods research to explore how users interacted with the support system.
User Interviews
Highlighted pain points in trust, past failures, and lack of clarity
Surveys
Surfaced consistent dissatisfaction with findability and usefulness
Analytics Review
Identified high bounce rates and failure loops in key flows
Heuristic Audit
Mapped misalignments between system labels and user mental models
One user summarized the sentiment well: “I tried going through the appropriate process, but at the end of the day I still had to call someone.”
Patterns pointed to a core insight: users weren’t bypassing the system because they were lazy, they did it because previous attempts to use the portal had failed them. They had learned the system couldn’t be trusted to deliver results, and resorted to workaround behaviors.
3. Strategy: Solving the Right Problem
We developed a strategic proposal to restructure the support experience at multiple levels. This proposal was organized into three progressive stages based on user needs and business readiness.
Update content taxonomy- Rebuild support content information architecture
- Re-map all support content
- Define governance procedures and content lifecycle
- Optimize search functionality
- Relaunch virtual agent
- Map end-to-end support delivery across channels
- Streamline backend flows
- Redesign front-end UX
4. Execution: Architecture, Design, and Alignment
Content Taxonomy & Information Architecture Redesign
We began with a comprehensive content audit, cataloguing over 500 support categories. We imported these into Mural, clustered them by theme, and analyzed labeling patterns and redundancies. Many categories overlapped in vague or inconsistent ways, undermining search and navigation.
We used tree testing via Optimal Workshop to evaluate the clarity and logic of reorganized taxonomies. We ran three rounds of testing with 65 participants, adjusting the taxonomy based on user input and failed tasks. As the structure improved, misclassification rates dropped significantly.
The final taxonomy was aligned to ServiceNow’s system capabilities and handed off to the Knowledge Management team for implementation.
UX & Visual Design
With the content groundwork in place, we turned our attention to the portal interface. Our primary design goal was to minimize dependency on support providers by improving self-service outcomes. We mapped new user flows based on our improved information architecture, aligning them to common user goals. Rapid iteration helped us quickly test low-fidelity prototypes.
Despite this, we encountered repeated breakdowns during developer handoff. While initial sprint reviews were positive, engineering later deemed our solutions too complex. This created delivery delays and eroded trust.
To resolve the issue, I restructured our design validation process:
Introduced mid-fidelity feasibility checkpoints to flag complexity earlier
Embedded system documentation into design artifacts
Facilitated cross-team retrospectives to rebuild rapport
These steps restored alignment and enabled us to deliver a feasible, validated interface redesign.
