Moon AI
Founder & Product Designer · October 2025 to present
Making an AI coach that people actually want to talk to.
CBT works. The clinical evidence is overwhelming. But access is the problem: a qualified CBT therapist costs £80 to £150 a session in the UK, NHS waiting lists stretch to months, and for most people it's simply not a realistic option. At the same time, a Harvard report found that millions of people are already turning to AI for emotional support and personal guidance, not because AI is ideal, but because it's available, non-judgmental, and always on. Moon's thesis: what if AI was taught to facilitate CBT properly? Not simulate a therapist, not generate affirmations, but apply the cognitive and behavioural techniques that make CBT effective (thought records, behavioural activation, cognitive restructuring) through conversation, at the moments people need it most. The MVP is focused on users who want to stop smoking and vaping.
I'm the sole founder. I've run the research, defined the product strategy, designed the end-to-end experience, and built a working prototype. Moon is a subscription product with a freemium tier and a corporate wellness benefit model. It's currently at prototype stage, with investor conversations ahead.
Before designing anything, I spent time understanding the category. I benchmarked Calm, Headspace, Fabulous, MyFitnessPal, Try Dry, Habitify, and Noom. I also researched CBT frameworks and techniques extensively, including interviews with three CBT practitioners to understand how the methodology translates outside a clinical setting. What I found shaped every major design decision that followed.
Two things stood out from the benchmarking. First, the visual direction. My initial exploration was bright, white, dashboard-heavy. The first version of Moon was actually called Momentum and it looked more like a fintech app than a wellness product. Benchmarking Headspace, Calm, and Fabulous pushed me toward calmer, darker tones and an illustrative icon style that added personality and made the product feel approachable rather than clinical. Second, the interaction model. Most wellness apps are built around logging: check in, record a mood, track a streak. My research showed that 63% of users had stopped using wellness apps because they felt impersonal. Logging was the wrong foundation. Conversation was the right one.
Research that built the product
I ran 10 research studies across UK, US, and Australian markets with 398 participants before committing to a product direction. The research was deliberately structured to validate the problem space before anything was built.
The studies:
- Building Better Habits survey (140 participants)
- Onboarding flow, moderated usability (10)
- Baseline setting, unmoderated usability (10)
- Habit Plans survey (18)
- Chat vs dashboard preference test, unmoderated (10)
- Habit plan interviews (10)
- Proposition validation survey (100)
- Willingness for AI coach survey (100)
- CBT techniques with practitioners, interviews (3)
The findings drove every major product decision. 84% of respondents had tried to build or break a habit in the past year. 71% said staying consistent was their biggest challenge, confirming the barrier is consistency, not motivation. 68% were comfortable receiving personalised AI-driven wellness guidance, and 74% said they'd be more likely to continue using an app if their AI coach felt unique to them. 81% expected AI wellness tools to be transparent and evidence-based, which validated the CBT-grounded positioning rather than a generic motivational approach.
The commercial signal was clear too. Three in four respondents said they'd try a new wellness app combining personal coaching with behavioural science, and 78% were willing to pay $5 or more per month.
63%
Stopped using wellness apps due to lack of personalisation
The finding that drove the conversational-first architecture
The pivot
The most consequential design decision was switching Moon's home screen from a logging dashboard to a conversation thread. The insight came from pushing on a single question: what does Moon actually know about a user after 30 days of tap-to-check-in data? The answer is almost nothing emotive. Just numerical signals: checked in or didn't, streak maintained or broken. After 30 days of conversation, Moon knows how you think, how you act, and how you feel during that period. That's the data that makes CBT interventions meaningful rather than decorative.
The concept was always that Moon would include conversational UI, but it was paired with conventional tracking inspired by Try Dry. My secondary research showed that users were already turning to AI for personal support, and my primary research confirmed it: interviewees were open to using AI as a coach, not just a tracker. That made me rethink the logging strategy entirely. If the whole interaction could be conversational, why split the experience between logging and talking? Just log in the chat.
74%
More likely to continue if their AI coach felt unique to them
Personalisation drove the conversational-first decision
The onboarding followed the same principle. My first version was long, modelled on Noom's screening process. In early moderated usability sessions I could see users losing engagement. I also knew there would need to be more than just intake questions: platform onboarding, interaction patterns, and the presentation of a bespoke habit plan. The solution was to reduce the visible onboarding to 6 core questions, then continue the rest as a conversation with Moon. Moon asks follow-up questions while building the user's personalised plan in the background. The user stays engaged because they're talking, not filling in a form. The platform learns more because conversation surfaces richer data than checkboxes.
Explore the prototype
A working prototype of Moon. Explore the onboarding, habit planning, and coaching experience.
Static prototype, AI coaching integration in progress.
What I learned
Moon is a different kind of portfolio project than BA or Agoda. It's not about leading a team or navigating an organisation. It's a demonstration of what happens when research, product thinking, and design execution sit in one pair of hands. I took an insight from a Harvard report, validated it with 398 people across three countries, benchmarked the category, defined a product thesis, designed the full experience, and built a working prototype. The lesson is that the skills I built leading large teams (research rigour, strategic framing, quality standards) translate directly to founding. They're the same skills applied at a different scale.
More projects
see all
See projects
see all