AI Health Companion
Wellness that actually fits your life — not a generic plan
A concept study exploring how AI could make preventive health personal, culturally aware, and genuinely usable for everyday people.
This is a concept study, not a product. Nothing has been built, and no users have been interviewed or tested. The personas and interview themes are hypothetical, written to structure the thinking. Claims about what features would achieve are assumptions to test.
Overview
The Premise
Many wellness apps tell you to eat chicken and broccoli and do a 5am run. They ignore that you might eat dal and roti, work a 9-hour desk shift, follow intermittent fasting for religious reasons, or simply have never exercised before. AI Health Companion is a product concept built around a different premise: that health guidance only works when it reflects how someone actually lives. This concept study documents the product thinking, proposed discovery approach, feature set, and delivery structure behind that idea.
Problem Space
What is broken today
Working assumptions behind the concept, not research findings.
One-size-fits-all health apps
Health apps often use the same templates regardless of cultural background, dietary norms, religion, or lifestyle. The assumption is that users disengage when advice feels irrelevant or unachievable.
No real-time physical feedback
Most apps give static exercise instructions with no live feedback. Poor form leads to injury, discouragement, and dropout — especially for beginners.
Reactive, not preventive
Existing tools log activity but do not connect lifestyle patterns to longer-term health trends. There is no guidance layer that says "you have been sedentary 5 days — here is what that means."
Product Vision
The guiding statement
"An AI health companion that learns your culture, lifestyle, and goals — then gives you guidance that is realistic, relevant, and respectful of how you actually live."
Target Users
Who this is for
- Desk-based professionals (25–45) with limited time and high sedentary risk
- First-generation immigrants and diaspora communities whose food culture is ignored by mainstream apps
- Health beginners who need encouragement, not intimidation
- Prevention-conscious adults who want to avoid chronic illness without drastic lifestyle change
- People with religious or cultural dietary requirements (halal, jain, kosher, religious fasting)
Personas
Hypothetical personas
Invented archetypes to frame the problem. They are not real people or research participants.
Priya, 32
Financial Analyst
Goals
- Lose weight without abandoning Indian food
- Build a sustainable routine around a desk job
Frustrations
- Apps recommend food she does not eat
- Generic advice triggers guilt not action
Needs
- South Asian meal suggestions
- Short desk-break routines
- No shame-based tone
Marcus, 26
Warehouse Operative
Goals
- Improve energy on shift patterns
- Eat better on a budget
Frustrations
- Wellness content targets flexible-schedule professionals
- No shift-worker acknowledgement
Needs
- Affordable simple meals
- Nudges that fit irregular hours
- Non-judgmental tone
Helen, 47
Marketing Manager
Goals
- Stay mobile and energetic
- Avoid conditions her parents had
Frustrations
- Apps target 20-somethings
- Hormonal and age-related changes are invisible
Needs
- Age-appropriate exercise guidance
- Preventive lifestyle nudges
- No overwhelming dashboards
Core Features
What the product does
| Feature | What it does | Why it matters to user | Expected product value (assumption) |
|---|---|---|---|
| Cultural Onboarding Engine | Captures background, religion, dietary requirements, location, and food habits | Advice feels relevant from day one | Intended to drive activation and trust; the main differentiator |
| AI Posture & Form Feedback | Uses camera to detect body position during exercises; flags issues in real time | Aims to reduce injury risk and build confidence | Expected to raise session completion and reduce dropout |
| Culturally Relevant Nutrition | Generates meal suggestions based on cultural food profile, not calorie templates | Suggestions use ingredients the user knows | Expected retention driver, if irrelevant advice is a main reason people abandon health apps (an assumption to test) |
| Work-Pattern Wellness Nudges | Desk stretches, posture resets, and hydration reminders based on work schedule | Small, regular prompts that fit the working day | Expected to drive frequent, daily engagement |
| Apple Health Connection (optional) | Read-only import of activity and sleep. No heart rate or other biometrics. The app works fully without it | Less manual logging for people who already use Apple Health | Reduces friction without making the product depend on one platform |
| Weekly Preventive Digest | Summarises patterns with one key insight | Awareness of trends before they become problems | Expected to support retention and perceived value |
| Adaptive AI Profile | Refines recommendations as behaviour is captured | Guidance improves the longer you use it | Intended to build a moat through switching cost |
Discovery
How the thinking was structured
- Desk research into health app dropout and ethnic differences in health risk (sources listed below)
- Hypothetical interview themes (no real interviews) mapped to 3 invented persona archetypes
- Jobs-to-be-done mapping: "When I [situation], I want to [motivation], so I can [outcome]"
- Assumption mapping: highest-risk assumptions identified, with low-fidelity tests proposed to check them (not yet run)
- RICE scoring applied across feature candidates to define MVP boundaries
- Ethical guardrails defined early: no medical diagnosis language, no prescriptive outcome claims
Sources
Published evidence read for this study. It supports two points: people often stop using health apps, and health risk differs by ethnic background. It does not show that culturally relevant advice improves retention. That remains the assumption this concept would need to test.
- Meyerowitz-Katz G, Ravi S, Arnolda L, Feng X, Maberly G, Astell-Burt T. Rates of Attrition and Dropout in App-Based Interventions for Chronic Disease: Systematic Review and Meta-Analysis. J Med Internet Res 2020;22(9):e20283.Pooled dropout of 43% (95% CI 29 to 57) across 17 studies of app-based interventions for chronic disease.
- Baumel A, Muench F, Edan S, Kane JM. Objective User Engagement With Mental Health Apps: Systematic Search and Panel-Based Usage Analysis. J Med Internet Res 2019;21(9):e14567.Median 30-day retention of 3.3% across 93 mental health apps. These are mental health apps, not fitness or nutrition apps, so it is indicative only.
- NICE. Overweight and obesity management (NG246), recommendation 1.9.11. 2025.Advises lower BMI thresholds (overweight from 23, obesity from 27.5) for people with a South Asian, Chinese, other Asian, Middle Eastern, Black African or African-Caribbean background, because cardiometabolic risk occurs at a lower BMI.
- Diabetes UK. Ethnicity and type 2 diabetes.States that Black African, African Caribbean and South Asian people are at risk of type 2 diabetes from age 25, compared with 40 for the white population.
- Ang SM, Chen J, Liew JH, et al. Efficacy of Interventions That Incorporate Mobile Apps in Facilitating Weight Loss and Health Behavior Change in the Asian Population: Systematic Review and Meta-analysis. J Med Internet Res 2021;23(11):e28185.Across 21 studies, app-based interventions produced a small to moderate weight-loss effect, and most of the apps were culturally adapted or locally developed. The review did not test whether cultural adaptation itself improves results.
MVP Scope
Proposed first release
In Scope
- Cultural onboarding flow
- Culturally aware meal suggestions
- AI posture feedback on selected bodyweight exercises
- Work-pattern wellness nudges
- Optional Apple Health read (activity and sleep only)
Out of Scope
- Heart rate and other biometric data
- Full wearable biometric analysis
- Mental health layer
- Community features
- Live nutritionist consultation
- Advanced diagnostics
MoSCoW
Prioritisation framework
Must
- Cultural onboarding
- Culturally aware nutrition engine
- Basic posture tracking
- Wellness nudges
- Optional Apple Health read (activity, sleep)
Should
- Personalised meal planning
- Progress dashboard
- Form correction library
- Adaptive difficulty
Could
- AI coach chat interface
- Social accountability features
- Gamification layer
Won't (v1)
- Medical diagnosis
- Guaranteed outcome claims
- Live consultation
User Journey
End-to-end flow
Onboarding
User answers 8–10 questions: location, background, religion, dietary requirements, lifestyle, work pattern, health goal, experience level.
Profile Built
AI generates initial profile: suggested meal style, exercise approach, nudge schedule.
First Workout
User films via camera; AI gives real-time posture feedback during a guided session.
Nutrition Feed
Personalised meal suggestions based on cultural profile — suggestions, not a rigid plan.
Daily Nudges
Desk stretch reminders, hydration prompts, and mobility flows timed to work pattern.
Weekly Digest
Activity and sleep (if Apple Health is connected), and one preventive insight based on patterns.
Adaptive Loop
AI refines suggestions as user feedback and behaviour is tracked over time.
Delivery Structure
Epics and stories
"As a new user, I want to answer questions about my cultural background and dietary requirements during setup, so the app understands my food habits from the start."
Acceptance Criteria
Onboarding captures country of origin, current location, religion (optional), dietary type, and 3 commonly eaten foods.
Profile generates first nutrition suggestions within the onboarding session.
"As a beginner, I want real-time feedback on my squat form using the camera, so I do not injure myself while learning."
Acceptance Criteria
Camera activates during flagged exercises.
AI detects at least 3 key form issues (knee cave, back rounding, insufficient depth).
Feedback appears as on-screen text within 2 seconds.
Session ends with a form summary.
Metrics
How success would be measured
North Star
Weekly Active Health Actions
Users completing 3+ meaningful interactions per week
Activation
Onboarding Completion Rate
% completing full cultural profile in first session
Engagement
Daily Nudge Response Rate
% of delivered nudges resulting in a user action
Retention
30-Day Return Rate
% returning and completing a health action 30 days after signup
Adherence Proxy
Meal Suggestion Acceptance Rate
% of suggestions saved or marked helpful within 7 days
Risks
Known risks and mitigations
| Risk | Impact | Mitigation |
|---|---|---|
| Camera permission and lighting dependency | High | Clear onboarding permission flow; graceful degradation to static tips if camera unavailable |
| Cultural nutrition data gaps at launch | Medium | Launch with high-prevalence dietary cultures (South Asian, Middle Eastern, East African); expand iteratively |
| Medical liability from health recommendations | High | Legal review of all copy; no diagnostic language anywhere; clear disclaimer on every guidance screen |
| Apple Health API changes | Low | Abstraction layer; the connection is optional and read-only, and the app works without it |
| User privacy concerns around camera usage | High | On-device ML processing; no footage stored or transmitted; explicit consent screen |
Ethical Disclaimer
Scope of guidance
The product would not provide medical diagnosis, clinical advice, or guaranteed health outcomes. All guidance is informational and based on general wellness research. Users with existing medical conditions should consult a qualified healthcare professional. Posture feedback is provided as general movement guidance only — not physiotherapy or clinical assessment.
Roadmap
Illustrative phased plan
Foundation
- Core onboarding
- Cultural nutrition engine
- Basic wellness nudges
- Optional Apple Health read (activity, sleep)
Movement
- Camera-based posture feedback
- Bodyweight exercise library
- Form correction alerts
- Progress tracking
Personalisation
- Adaptive AI model
- Meal planning depth
- Weekly digest
- User feedback loop
Expansion
- Wearables beyond Apple Health
- Community layer
- Expanded exercise library
- Coach interface
PO Reflection
What I learned from this concept study
The most interesting constraint in this concept is the ethical boundary. The product must feel genuinely helpful without ever overstepping into medical advice. Defining that line — in copy, onboarding, and feature design — was as important as the features themselves.
My hypothesis is that culturally aware personalisation is under-served in mainstream health apps, many of which default to western dietary models because it is simpler. This concept challenges that default. Whether there is a real underserved market here is the first thing discovery would need to test.
The technical complexity of real-time posture tracking is genuine, but the harder problem is trust — getting users to grant camera access and believe the guidance is for them. That is a discovery and design challenge long before it is an engineering one.