SEE WORK BELOW ↓ SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓SEE WORK BELOW ↓
How intelligent search and recommendations transformed HelpZone into a self-serve support engine.
Please look on your desktop to see this case study :)
Team: Moon Sun (Lead Product Designer) | Liz Alexander (Content Designer) | Sarah Engleman (UX Researcher Lead) | , Hyong Woo Hahm (UX Researcher) | Stan Cole (Engineering Lead/Manager) | Arlene Ccheng (PM)
One-liner
I led the redesign of HelpZone, Cisco's internal support platform, using AI and behavioral data to inform key product decisions.
I led the redesign of HelpZone, Cisco's internal support platform, using AI and behavioral data to inform key product decisions.
Outcome
Drove measurable outcomes across the board: 85% of employees attempting self-help, 80% tile adoption, 70% self-resolution confidence, 12% improvement in employee satisfaction, and 50% gains in tracking efficiency.
Drove measurable outcomes across the board: 85% of employees attempting self-help, 80% tile adoption, 70% self-resolution confidence, 12% improvement in employee satisfaction, and 50% gains in tracking efficiency.
My role
I led this project end to end as the sole Lead Product Designer, defining the problem, setting the direction, and delivering the work in collaboration with Product, Content, Research, and Engineering.
I led this project end to end as the sole Lead Product Designer, defining the problem, setting the direction, and delivering the work in collaboration with Product, Content, Research, and Engineering.
Context
HelpZone is Cisco's internal employee support platform, home to a knowledge base spanning IT and People & Culture (HR) resources, with options to open cases or reach live support. Despite its breadth, more than 75% of employees bypassed it entirely, turning to colleagues or direct support channels instead. The platform had a trust problem - and with it, a self-service problem
Problem
Entry-point feedback surfaced a recurring theme across users: the pathways to live support and case submission were too hard to find. Employees weren't just struggling with self-service, they couldn't reach human support either, leaving them with nowhere to turn.
Working with UX Research, I ran a 20-user study to pinpoint why support channels felt invisible. We asked participants to open a case, and 16 out of 20 didn't scroll far enough to even find it. The channels weren't missing. They were just out of sight.
The fix seemed obvious: surface support channels above the fold. But stakeholders pushed back. Live agent support was being cut by two-thirds, and the business couldn't absorb a spike in opened cases or calls — they needed those channels hidden, not highlighted.
Initial idea of surfacing the support channels above the fold
Initiating the Process
To balance user needs with business constraints, we crafted 'How Might We' statements to keep our exploration grounded and focused.
How can we help users find their solution efficiently while encouraging self-help?
How do we get users to self-help before looking for a live agent?
How do we help users get to where they need.
Constraints
Design Thinking
At this time, Cisco's internal AI tool called Circuit was introduced as a internal tool for employees to use in the workforce. I hosted a workshop with my team to understand how we could possibly leverage this AI tool to encourage users to do self help.
One of the proposals was using AI to show options in the typeahead as well as as summarizing articles. This rang a lightbulb. After that, I used AI to generate more ideas on where we could possible incorporate AI into HelpZone.
PowerBI showing data where users are clicking on the homepage of HelpZone
Embedded CircuIT in the typeahead to show suggestions to users while they are typing
Most support platforms treat search as a starting point, here are ten articles, good luck. CircuIT flips that. It reads the user's query, generates a summarized solution, and pairs it with the most relevant articles. The goal: resolve the issue right there in the search results, before the user ever needs to open an article or contact an agent.
Data-Informed Design
Even with these changes, business requirements kept support channels off the homepage - but users made it clear they wanted a way to reach a live agent there. So we looked at the data: fewer than 1% of users clicked the homepage's "Featured Articles" section, since the content wasn't personalized. Yet IT operations still required it, taking up real estate that could serve a better purpose.
I proposed pulling a component from Bridge, one of our other employee platforms: the Spotlight modal. Reusing components was always an easy yes internally, less to build, and it kept our employee platforms consistent. The idea was to use the modal for "Featured articles," with a discreet "contact support" link pointing to an article with phone numbers. I also worked with engineering on personalizing the articles themselves, and we landed on surfacing them based on the user's location.
So although the tradeoff was that we could not surface the support channels like the way user wanted, they were still able to access some kind of live support if it was really needed.
We used PowerBI behavioral data to determine which categories employees were searching for most. That data directly informed which tiles appeared on the homepage and in what order, the most-used categories were surfaced first, so employees could find what they needed without scrolling or guessing.
Each category featured a curated list of the most reliable, up-to-date resources maintained by the organization team. PowerBI also revealed that employees overwhelmingly used the subcategory filter over other filter options, so instead of presenting a long, flat list of articles, we organized the curated content into subcategory groups, making it scannable and easier to navigate.
Building trust with users while meeting business needs
Building trust with users while meeting business needs
We couldn't put support channels on the homepage — but we could make them reliably present everywhere a user goes after that. By surfacing live agent access the moment self-help begins, we solved both sides: the business kept homepage escalation low, and users always knew help was within reach.
We couldn't put support channels on the homepage — but we could make them reliably present everywhere a user goes after that. By surfacing live agent access the moment self-help begins, we solved both sides: the business kept homepage escalation low, and users always knew help was within reach.
Smart Routing, Seamless Experience
Smart Routing, Seamless Experience
The organizational split between IT and P&C was a backend reality, not something users should have to navigate. We solved this by using page context to determine which support channels appeared. An employee browsing IT content only sees IT support options. An employee on a P&C page only sees P&C options. The experience feels unified, but every case lands with the right team
The organizational split between IT and P&C was a backend reality, not something users should have to navigate. We solved this by using page context to determine which support channels appeared. An employee browsing IT content only sees IT support options. An employee on a P&C page only sees P&C options. The experience feels unified, but every case lands with the right team
Impact
The organizational split between IT and P&C was a backend reality, not something users should have to navigate. We solved this by using page context to determine which support channels appeared. An employee browsing IT content only sees IT support options. An employee on a P&C page only sees P&C options. The experience feels unified, but every case lands with the right team
increase in self-resolution user confidence
lift in tile adoption compared to the quarter prior.
improvement in case tracking efficiency, cutting the time agents spent rerouting misplaced cases
of employees now attempt self-help before contacting an agent, up from just 36%. A 49-point shift that proved trust in the platform was earned