Project: Koodo Traditional Search to GenAI-Powered Search
Client: Koodo Mobile
Platform: Help Center (https://www.koodomobile.com/en/help)
Role: Principal product designer
Team: Principal product designer, Product Management, Engineering & Dev team, Content Strategist, Legal team

PROBLEM STATAMENT

Koodo’s traditional help page saw frequent user drop-offs and repetitive search queries that didn’t convert into meaningful resolutions.

Users often got overwhelmed with the volume of static FAQs or couldn’t find what they needed quickly enough.

Koodo needed a smarter, scalable solution to improve self-service and reduce support calls.

Discovery & Research

Empathize

To start, we needed to understand:

Who are our users?
Mostly mobile subscribers aged 18–55 with a range of tech-savviness levels.
What are their pain points?
- Frustration with navigating dense FAQ content
- Repetition in searches (e.g., “How to reset voicemail”)
- Long resolution time

Methods Used:

User Session Recordings (Hotjar/Contentsquare): Tracked where users abandoned the page.
User Surveys: Asked real users how satisfied they were with the help experience.
Heuristic Evaluation: Found gaps in search interaction, result relevance, and readability.

Define the Problem

“How might we enable users to find answers more quickly and intuitively without overwhelming them with static, outdated content?”

We defined our success criteria (Set KPI’s):

15–20% reduction in help tickets
Improve Call containment on Help page by 10%
Enhance time-to-resolution under 15 minutes

Ideation & Strategy

Key Design Goals:

Conversational Simplicity – Use GenAI to mimic human-like interactions
Trust and Transparency – Indicate AI involvement and offer fallback options
Efficiency over Exhaustiveness – Serve refined, relevant answers without long text blocks

Solutions Explored:

Generative AI chat-like flow vs. traditional filtered search
Contextual microcopy for error states and suggestions
Escalation path to live chat [Koodoo assist or KEM] if GenAI fails to resolve query

Design & Prototyping

Wireframes & Flows

I created low- and mid-fidelity wireframes to experiment with:

Quick GenAI summary capabilities
Sticky search input at the top
Immediate auto-suggestions
Follow-up question capability to refine results

Key UX Enhancements:

AI prompt clarity: Reframed “Ask us anything” into action-based prompts like “What can we help you with?”
Result hierarchy: Most likely answers shown first, dynamic loading for less relevant ones
Conversational response format: Clean, humanized tone of voice and enable easy screen reader for accessibility

Validation & Testing

Methods:

Usability Testing (5 sessions):
Users were asked to find answers to real scenarios (e.g., “How to change my plan”). We monitored time-on-task and success rate.
A/B Testing:
Compared the legacy Help search vs. GenAI search on key metrics like:
- Completion rate
- Engagement with follow-up questions
- Bounce rate
Content QA:
Ensured AI-generated responses matched Koodo’s brand tone and were legally compliant.

Implementation & Collaboration

Partnered closely with engineering, content and legal team to:

Implement token limits and safeguard prompts
Add smart fallback logic to connect users to agents when AI confidence was low
Ensure WCAG 2.1 accessibility in all interface elements

Results & Impact

Within the first 90 days of launch:

✅ 13% drop in help tickets submitted through contact form
✅ 18% increase in Help page satisfaction score (measured via post-interaction survey)
✅ Time-to-resolution dropped from 33 minutes to 13 minutes
✅ 3x engagement with Help search bar (mobile and desktop)

WHAT’S NEXT?