AI Help Center
Client:
Study Design:
My Role:
Date
Location
Eventbrite
Qualitative multiphase design
Lead UX Researcher
September-October 2025
Remote
Team:
8 person team with 1 skip manager, 3 PMs, 1 Designer, 1 Analytics partner, 1 Voc, and 1 Lead Researcher (me)
Executive Summary
Users struggled with a help center that offered repetitive, unhelpful responses and slow human support. This research explored how AI could be integrated to resolve issues faster and reduce reliance on human agents.
BACKGROUND
User Problem & Key Challenges
User Problem: Users were increasingly frustrated with a help center that did not provide quick answers, a chatbot that gave repetitive useless answers and human support that took a long time to engage with no guarantee of a solution.
Challenge: The team was pushing hard for AI implementation without proper scaling or planning. Their research questions were 1 dimensional and I embedded additional research that would advocate for a careful and strategic implementation.
Research Questions & Hypothesis
Team Question:
1. How can we incentivize users to resolve issues using non-human support methods (e.g., AI chat, help center articles)?
2. What factors cause users to seek human agents?
3. What is the ideal level of AI involvement users are comfortable with?
My Questions:
1. how are other dual sided marketplaces implementing AI?
2. Where should we prioritize human support vs. AI support?
3. When does AI break user trust?
Hypothesis: If we implement AI into our help center, then we will see a reduction of human support requests because users will have the ability to solve common issues on their own.
METHODS
Participants
Target Sample:
Event creators with recent experience in the Eventbrite help center.
Recruitment:
10 users were recruited through random selection of internal data from analytics.
Study Design
Internal Audit
Although the team was aware of user frusteration in the help center, they had not explicitly pinpointed the exact user problems.
Competitive Analysis
It was important to understand marketplace standards though as to enter at a competitive level. Additionally, this would provide ideas of strategic implementation.
Interviews
Users were asked to complete tasks on desktop and mobile view to test designs. We also gathered sentiments around AI during this phase
Analytics & Presentation
Analytics
Voc and customer support feedback was synthesized along with results from the competative analysis and interviews.
Presentation
2 presentations were made since timelines were tight and the designer needed preliminary findings to work with.
RESULTS
Findings
1. The competitive analysis revealed that other marketplaces integrated AI support based on the complexity of the user issue (more complex issues were automatically handed to human support)
2. Users utilized the solution method they perceived to be the fastest and appreciated a multimodal approach. User trust was broken when expectations did not meet reality but expectations for an AI chatbot were hard to discern.
3. Users were open to any AI implementation as long as there was transparency around data gathering and that it truly solved their issues.
Impact
The team reframed their approach on who was offered AI vs. human support (originally, higher tiers had access to human agents). Their new approach focused on the complexity of the issue guaranteeing less friction for the users.
The team also moved away from an "AI first" help center to an incremental implementation where AI would be embedded only in issues we were confident it could solve.
Lastly, the team recognized there needed to be a help center revamp that would rethink how we present articles, where we embed AI and how we can swiftly transition the user to human support and effectively reduce redundancy.
REFLECTIONS
Lessons
Since conducting this project, my knowledge on AI (it's limitations and strengths) has expanded immensely. I am proud that I slowed down the implementation into phases hat wouldn't break user trust. But I would now go one step further and work with the designer to come up with creative AI solutions that would go beyond an AI chatbot (the market is highly saturated with these). Additionally, I've now studied good "AI Design" that can facilitate the adoption of these AI features for users and simultaneously, truly solve real user problems (AI chats are perceived as bandaid solutions).
I knew even back then we could be more creative with how we implemented AI
examples from the competitive analysis demonstrating what other solution methods are offered
although designer was hesitant, the data gathered demonstrated that some design changes I advocated for increased user navigation
I knew even back then we could be more creative with how we implemented AI