PrOJECT OVErVIEW
Redesign Cisco’s internal HelpZone leveraging AI to help employees quickly resolve issues, reduce agent dependency, and maintain productivity.
Project type
product design & strategy
YEAR
2024-2025
MY ROLE
lead designer
Company
cisco
Problem
Due to reduced investment in support agents, Cisco’s internal HelpZone experienced longer wait times, slowing employees’ ability to resolve issues and stay productive. Employees needed a faster, self-service way to find solutions without relying on agent support
Solution
We introduced an AI-powered HelpZone experience that guides employees to relevant solutions through intelligent search and recommendations, enabling faster issue resolution, reduced support wait times, and less reliance on human agents.
of employees attempted self service on Help ZOne
more users reported having no difficulties in navigating HelpZone compared to the previous quarter
AI suggested search prompts
Our enhanced search functionality prioritizes CIRCUIT responses for immediate answers to your questions while also giving you seamless access to additional support resources.
ai suggested search results and recommendations
Circuit Provides a generated AI response based on what the user searches for. It also generates suggestions and recommendations for the user to view
Homepage Spotlight section
This area highlights important updates, featured resources, and links to support options, including in-person assistance.
New Contact Support experience
To encourage self-service while still supporting urgent needs, we designed a tiered support access model. For urgent P&C, IT, or Facilities issues, a discreet link was placed on the highlights card, directing users to a dedicated page with support phone numbers for quick access. This reduced visual prominence nudged users toward self-help options before escalating to live agent support.
Additionally, a consistent “Contact Support” CTA was placed in the top-right corner across all HelpZone pages, providing access to a drawer of support channels. This ensured support remained easily discoverable without interrupting the primary self-service experience.
BAckground
I joined Cisco as part of the HelpZone product team and worked on three MVP launches. Cisco’s original internal support experience was often unintuitive, slowing employee productivity. User research consistently showed that employees struggled to find support channels, and while this insight was initially deprioritized due to concerns about agent capacity, continued research led stakeholders to reconsider.
Amid layoffs impacting support teams, self-service became a priority. This led us to integrate AI into the HelpZone experience to encourage employees to resolve issues independently before reaching out to support.
original design before 12/2022
Research & inspiration taken
we kickstarted this project by conducting user interviews with (at the time) current support and looking into what other companies were doing with their own support experience.
Homepage
For the first MVP launch we prioritized on how to provide users quick access to articles.
For our first map launch we introduced the top 5 main topic navigation (based on user research on what was most requested) on the top where users can navigate thru different main topics to be introduced to a side sub navigation that will hence show a curated list of updated articles for each subcategory
Chat support was also available for the first time on homepage
we used power BI to understand where users were clicking on the most on the homepage. We found that most users were clicking on the search by a long shot over everything else on the page
MVP launch 02 | OCT.2023
Homepage
For the second launch we decided to get rid of the side subtopic navigation and have users click on to a curated list of articles from their main topic selection. We decided to do this because on our previous user interview studies, most users felt like the first mVP was still "visually" too much. We tried to approach our next MVP by creating a more simplified design.
Also by request from our stakeholders we included "featured articles" that will change based on time.
We used Power BI again to see where users were clicking on the most. Again, search bar was most commonly used. Less than 0.5% of the clicks on the homepage was featured articles. We used this data to argue that "featured articles" were taking up space on the homepage. Based on the user interviews we did at this time, most users have also complained about the visibility of support channels.
Constraints & Challenges
One of the biggest challenges in HelpZone was an ineffective search experience. Due to Cisco’s long history, the platform contained many outdated or irrelevant articles that had not been properly updated or archived, making it difficult for users to find accurate information. Although support agents were actively improving content quality, users still struggled to trust HelpZone as a reliable source.
Additionally, stakeholders remained cautious about surfacing support channels too prominently, preferring to limit access to agent support unless users had urgent needs. There was also a requirement to continue highlighting certain articles from previous launches, even though user research showed they had low engagement. These constraints shaped how we approached improving search, trust, and self-service within HelpZone.
search constraints
From a UX perspective, the previous typeahead experience was well designed. It visually consistent with the intranet, intuitive, and widely used. However, there are opportunities to improve the technical implementation to increase recommendation accuracy.
The problem with the search was that In some cases, queries surface services instead of relevant articles, even when strong article matches exist. This suggested limitations in content matching. Addressing this while continuing to prioritize AI remains aligned with our strategic direction and goal of improving search.
We also recommended exploring popular searches, a common intranet pattern. As an MVP, this could surface globally popular queries related to the current input, with a longer-term path toward AI-driven, role- and location-based personalization.
