Transforming Self-Serve Support with myOneTrust CoPilot
myOneTrust was rich in knowledge but difficult to navigate. I designed the CoPilot experience to turn thousands of technical documents into clear, contextual guidance — helping customers find accurate answers faster, reducing support load, and improving self-serve discovery within a complex ecosystem.
Principal Product Designer
Jan 2025 - Mar 2025
PM, Engineering
Web
How I First Understood the Problem
When I began exploring the state of self-serve support on myOneTrust, I didn’t start with the UI.
I started by listening.
In surveys, interviews, and community posts, customers used words like:
“Byzantine.”
“Confusing as hell.”
“Worst support system ever.”
And yet, these same customers visited the portal every day.
Not because the experience was good — but because they had no other reliable path to answers.
myOneTrust had nearly 10,000 knowledge articles,
yet users often spent 15+ minutes searching for something as simple as a password reset.
The more content we produced to help them,
the harder it became for them to find anything.
That paradox was the root of the problem.
Why This Problem Mattered
myOneTrust isn’t just a documentation library —
it is the primary support surface for thousands of customers.
When users fail to find answers:
frustration spikes,
support tickets increase,
customer trust erodes,
and onboarding slows down.
Support teams felt this pressure in volume.
Product felt it through misrouted questions.
Customers felt it through repeated dead ends.
The cost wasn’t just operational —
it damaged the perception of the platform itself.
Our job wasn’t to build a chatbot.
Our job was to restore clarity and confidence in the entire support ecosystem.
35%
Improved onboarding process
25%
Increase in user retention
84%
Increase in time spent on website
My Design Approach
As Principal Designer, I had to consider not just UI patterns,
but the psychology of trust, the complexity of AI behavior,
and the guardrails necessary in a highly sensitive privacy domain.
Prioritize trust over intelligence: Users must be able to trust that:
every answer comes from staff-written articles
nothing is hallucinated
citations are clearly visible
disclaimers are transparent
tone stays consistent with the myOneTrust brand
Trust isn’t a feature — it’s an infrastructure.
Reduce friction, not human interaction: The goal was not to deflect customers away from humans.
The goal was:
faster discovery
fewer barriers
clearer next steps
reduced cognitive burden
This meant designing off-ramps intentionally:
“Still need help?” → create a support ticket
“Want advice from peers?” → ask in Groups
“Need related topics?” → explore more articles
AI needed to support, not replace, community behaviors.
Align the AI experience with how people ask questions: People do not type like documentation authors.
They type like humans:
“reset password”
“videos not loading”
“cookie banner broken”
So CoPilot needed to interpret:
keywords
partial phrases
typos
incomplete thoughts
vague intent
The UX challenge was to design for low-effort entry and high-quality return.
Create a scalable knowledge pipeline: One of the biggest design responsibilities was shaping the content feedback loop:
upvotes → validate quality
downvotes → flag gaps
missing content → routed to technical writers
ambiguous answers → prompt refinement
recurring questions → tuning top 100–300 intents
UX wasn’t just about the chat interface —
it was about designing the system that keeps it alive.
The New Experience
The CoPilot experience is simple and intentional:
User logs in
Opens CoPilot from the top bar
Asks a natural-language question
Receives a concise answer + clear citations
Upvotes/downvotes the quality
If unresolved → guided to the right human or resource
CoPilot never pretends to know what it cannot prove.
It always reveals its source of truth.
This was a fundamental principle of trustworthy AI.
What Users Told Us
Internal testers — CSMs, Support, SMEs — said:
“I can finally find the right article in seconds.”
“This will dramatically reduce ticket noise.”
“Robust answers with the right context and citations.”
The sentiment was consistent:
AI wasn’t replacing humans —
it was removing the friction that kept users from reaching them effectively.
The Impact
Quantitative Targets
+4% improvement in “myOneTrust is easy to use”
–5% support ticket volume reduction
≥85% correct-answer rate across engagements
Qualitative Gains
Faster onboarding for new customers
Reduced “search fatigue”
Renewed trust in documentation quality
Improved community engagement (protected from AI dilution)
Stronger foundation for cross-platform AI initiatives
This project wasn’t just about answers —
it was about making myOneTrust feel navigable again.