Director’s cut · Responsible AI · 2026
Breakmate, the long version
People did not need a better listener. They needed a shorter path to someone who is one.
The evidence behind the problem and how we narrowed it. Then each call we made with what it cost, and what testing changed.
Role
Product Manager
Team
4: product and design, research, brand, prototype
Timeline
Jul – Aug 2026
Outcome
Tested prototype, 29 screens
7 min read · main cut 4 min

Light and dark mode on the same 2 screens: the patterns dashboard and the guided pause.
10
AI users from our recruiting screener
3
more calls, each with what we cut
6
fixes from 2 rounds of testing
170+
hours in the BRIDGEGOOD UX Design Apprenticeship
A Process
We didn’t start with “build an AI wellness app”
The brief was broad: how does conversational AI shape people’s judgment? My job was to turn that into a problem a 4-person team could own.
v1
How does AI shape judgment in uncertain moments? Too wide to design for.
v2
Frequent users seek relief from AI. Reliance shifts from the person to the chatbot.
v3
Reassurance quietly replaces self-trust, and today’s tools can’t recognize the difference.
How might we help young adults who turn to AI during moments of uncertainty build trust in their own judgment?
Our “how might we”

Our problem statement, rewritten on the board as it narrowed

Sketches by me and the team. A traffic-light signal and a conversation bridge fed the final concept.

Competitive analysis of assistants and wellness chatbots. Nothing tracked patterns across tools.

Working through the prototype together at BRIDGEGOOD with an Adobe Senior Product Designer

With my team, JMAC: Jaslyn Brown, Chris Cornelious and Annie Chan.
B Research
What we heard, and what it changed
80%
use 2 or more AI assistants (8 of 10), so no single app sees the whole loop
70%
had used an AI chatbot for emotional support (7 of 10)
70%
use AI every day or more (7 of 10)
From our recruiting screener: 10 AI users who applied to be interviewed. We recruited people who use AI for personal support, so this describes them, not all AI users.
Insight 1
Repeat questions are the signal, not screen time
People brought AI the small, recurring worries they’d be embarrassed to share with friends. It’s free and instant, and it “doesn’t need a background check.” The repetition, not the hours, is what signals reliance.
Insight 2
Detection isn’t the hard part. The hand-off is.
Even when a tool spots distress, very few people follow through to a human. The path to support has to be short and chosen by the user.
Insight 3
No single AI app sees the whole pattern
People split time across several assistants. Safeguards like ChatGPT’s trusted contact only see what happens inside ChatGPT.
The wider evidence
OpenAI · October 2025
1M+
people a week have ChatGPT conversations showing signs of suicidal planning or intent, by OpenAI’s own estimate.
Wysa · 19,000 users · 2024
2.4%
of users in crisis called a helpline, even when the app kept encouraging it. Its AI had detected 82% of those crises.
UBC · 296 students · 2026
Only people
reduced loneliness over 2 weeks. Chatting with an AI did no better than journaling. Texting a stranger did.
C Interviews
People feel the pull, and the gap
Content note · crisis language
The conversation that started our pitch
A real, anonymized conversation from our research, described here rather than shown. The user was in crisis. The assistant misread it: it agreed where it should have questioned and filled in assumptions the user never made, even contradicting them about their own words. It sounded supportive the whole time.
We opened the pitch with it, behind a content warning, because it showed the failure more clearly than any statistic could.
What we heard
One job seeker turns to AI when her frustration peaks. She was surprised how well it mimics empathy, but uneasy knowing it’s a machine and unsure the relief lasts. She compared it to scrolling: something we do, but probably shouldn’t.
A human-factors specialist who uses AI deliberately said she likes it agreeing with her when she tests a simple idea, and prefers it handing her options over working through her own. Without it, she’d feel less capable. Nothing looks wrong, and the judgment still moves to the AI.
Another interviewee brings AI the worries that feel too small for his friends. But whenever a real person is available, a friend or therapy his insurance covers, he picks the person.
That last pattern set our direction. Breakmate doesn’t try to be a better listener. It points people toward someone who is.
D More decisions
3 more calls, and what they cost
The main story keeps the 3 that shaped the product most. These shaped how it behaves.
Step in automatically, or let the user decide?
User-set limits, plus a direct line to someone the user picks.
Automatic blocking or escalation. It takes away the control we wanted to give back.
How much conversation data do we need?
Pattern analysis on the device only; conversations are never stored; disconnect anytime.
Cloud analysis of chat content. It might be more accurate, but it breaks trust. (Access to other apps’ data still needs to be validated.)
Our Adobe advisor asked: how do we make relying less on AI actually stick?
A values quiz in onboarding, so nudges and exercises point back to what each person cares about.
1 set of tips for everyone. Advice that isn’t about you is easy to ignore.
What we cut from version 1
A screen for managing every AI connection
I wireframed a place to connect or disconnect each assistant, with tools we couldn’t support yet marked “coming soon.” It wasn’t the core feature, and we had limited time, so we spent that time on loop detection and the guided pause instead.
The same privacy promise, “analysis runs on your device only,” lives on the consent screen in the final prototype.

Early wireframe, not in the final prototype. Tool names and the product name were updated for this page.
E Testing, round by round
What each round found, and what we fixed
Round 1 · our first idea
A postcard to your future self
At the center was a postcard: a note about a decision, sent to your future self and returned later to interrupt the reassurance loop.
4 of 5 couldn’t say what the postcard did until we explained it, and even then it didn’t feel like it would change anything. The fifth liked its playful, game-like feel. All 5 liked the goal; the postcard just wasn’t the way to get there.
Round 2 · after the redesign
A loop alert and a way back to people
We rebuilt around what testers could act on: a loop alert with app limits they set themselves, then exercises matched to their values and a hand-off to a real person.
All 3 called it straightforward. None got lost, so they hunted for flow problems and found 2: the dashboard jumped back to the top when you returned to it, and the values quiz let you continue with 1 pick when it asked for 2 or 3 (both are fixed in the final prototype). They liked the look and the welcome note, and the analytics gave just enough to dig deeper. Most of all, they weren’t left alone with the numbers. Exercises and nudges sat beside them, along with a way to reach people they trust.
01
Onboarding now says what Breakmate does up front. A tester got lost navigating because nothing explained the app’s purpose.
02
Selected states are more visible. Testers asked for active selections to be clearly visible.
03
The dashboard shows less at once. Testers liked the dashboard for accountability, but found too much on 1 page.
04
Back buttons moved below the Dynamic Island. Back-button icons overlapped the iPhone’s Dynamic Island.
05
The dashboard keeps your place. It used to jump back to the top when you returned to it.
06
The values quiz waits for 2 or 3 picks. It used to let testers continue with 1.
Qualitative notes. The team didn’t record timings or task completion, so there’s no task-success rate. A timed retest of the final flow is the next step.
F Design system + accessibility
Designed end to end, for hard moments
I designed every screen of Breakmate. People open it when they’re stressed, often late at night, so accessibility was part of the system from the start, not a pass at the end.
Mode
Body text
Button
Muted text
Light
12.7:1
5.4:1
5.3:1
Dark
12.5:1
7.4:1
6.3:1
Each ratio is the lower of the 2 surfaces the text sits on, background and card. WCAG AA asks for 4.5:1 on body text.
What I designed
The full product
All UI and interaction design, from lo-fi wireframes to the final prototype
A 163-variable design system with light and dark modes
The components, and the micro-animations testers praised
Brand and logo by Chris Cornelious
Accessibility
What I built in
Contrast checked on every screen
Minimum text sizes, system-wide
Tap targets of at least 44 points
Dark mode for late-night use
VoiceOver labels for every control
Reduced-motion alternatives
Back buttons below the island
Why it matters here
Stress is the use case
Small text and low contrast fail people first when they’re stressed or tired, so these were requirements, not polish.
G Pitch day
The crowd’s favorite part came from an advisor’s challenge
The capstone ended with a live pitch and demo at BRIDGEGOOD, closing out a UX Design Apprenticeship of more than 170 hours.
“Brilliant.”
Audience feedback
From a cybersecurity professional in the audience, who also pointed to universities as the right way in, matching the institutional model we had proposed.
Our advisors from Microsoft and Adobe saw it helping a much wider range of people than students.
The values quiz, added after our Adobe advisor’s challenge, was the audience’s favorite part of the demo.

Chris and Annie walking the room through the key features.

Receiving the UX Design Apprenticeship certificate.

The UX Design Apprenticeship certificate.

The BRIDGEGOOD apprenticeship cohort.
H Next steps
What comes next, and how we’d know it works
Proposed
What comes next
A first institutional pilot, such as a university counseling center
Deeper testing of the pause and discover flows
In-chat detection
A technical spike to confirm cross-app data access
Metrics
How we’d measure success
Users can explain the patterns we show them
Guided-pause completion rate
Self-reported sense of agency, before vs. after
Support taps that turn into real contact
False-positive rate of loop detection
Still open
Access and privacy
Big AI apps want more use, not less, so they have little reason to let us in. Privacy and browser workarounds were open questions too, and so was telling work chats from personal ones.
I What I learned
2 things I’ll carry forward
Research has to be the tiebreaker
Everyone arrived with a favorite solution, and some design calls turned into real conflict. What worked was making the research the referee: if a finding didn’t support an idea, it didn’t ship. I also learned to match work to people’s strengths and to check quality early, rather than step in once a deadline had slipped.
It isn’t a young-person problem
We scoped for young adults, but the pattern showed up across generations, rooted in loneliness, social skills, upbringing, emotional state and AI literacy. And a hand-off only works if the help on the other side is affordable and still there when the person comes back.
Want the short version?
The case study
The 4-minute cut has the 3 decisions that shaped the product and the flow that shipped.
Keep reading · case study 1 of 5
Product Strategy · Enterprise 3D · 2026
ATELIA
5 days from a blank brief to a pitch to Dassault’s leadership
© 2026 Moe Htet
Los Angeles