Fit Fuel: An AI-Augmented Mobile App
Problem
As the focal point of a four-week AI for UX Design course through Designlads, the project addressed the challenge of using an array of AI toolsets to create a new fitness app that fills gaps and white space in the current app marketplace. Current fitness apps adapt to biometrics, not to real life — and none treat coming back after a break as anything but a failure.
Challenge
Design for three very different audiences (busy professionals, lapsed returners, access-needs users) without shame mechanics or excluding the highest-risk group.
Approach
Market Research | Competitive Assessment | Ethics & Bias Audit | Persona Development | Journey Mapping | Design Brief | Prototype Creation | Accessibility Audit | Usability Study
Result
The FitFuel project validated lapse recovery as a genuine market white space — no competitor product treats returning after a break as anything other than failure — and translated that insight into a shame-free re-entry flow and onboarding prototype grounded in interview and market research. The final output created a sound app architecture, needing only some basic validations for readiness to take to market.
Reseach
Utilizing Claude AI, I identified that market data reframed the project around retention (industry 30-day retention: 3–8%). Competitive audit of 8 apps (Fitbod, Freeletics, Future, Peloton IQ, Ray, Whoop Coach, and others) confirmed AI is everywhere but adapts to biometrics only — never to calendar or life context. Interviews confirmed two dominant patterns: plan-life mismatch and rejection of shame mechanics via streaks and guilt prompts.
Deliverables:
Market Landscape | Interview Thematic Clusters | Trainer Synthesis | Source Validation Log
Design Process
I established the working rules before research began by identifying which AI tool does which job, I developed bias-audit checkpoints for every persona/synthesis output, and I created source-validation standards so every claim is labeled validated, directional, or hypothesis.
Deliverables:
AI Stack Brief | Ethics & Bias Auditing Brief | Prompt Library
Personas & Flows
I developed progressively advanced wireframes beginning with the establishment of page-level features and content to inform a design exploration process and evolving to a full design prototype that was the subject of a usability test. I ran a 12-user, 60-minute test focused on taxonomy, success of browsing to specific products, and store locator function to inform improvements to overall usability of the site.
Deliverables: Market Landscape Brief | User Interview Report | User Personas | Journey Maps | Design Assumptions Brief
Design
Working with Figma Make as the AI design engine, I built two flows to prove out and test the research hypothesis. Lapse recovery profile and the new-user onboarding avoid forcing lapsed athletes into a false "beginner vs. advanced" choice. The re-entry flow is three quick taps with no gap-shaming language anywhere, injury modifications shown up front. Each of these decisions was traced to a specific research or interview finding.
Key screens:
Flow 1: Landing → Account → Profile → Plan Delivery
Flow 2: Re-Entry Prompt → AI-Rebuilt Plan → Adapted Plan
Testing
Through both manual audits and development of specific Claude AI test prompts, I identified that the accessibility audit found two blocking issues: no keyboard navigation on six screens and no visible button focus state. Each issue was noted as a show-stopper and must be fixed before real users touch it.
Also, I ran an AI-driven usability scenario; the synthesis of that research found the plan doesn't visibly reflect user inputs, plus a flawed fitness-background taxonomy and aspirational schedule inputs. Both needed further design exploration and retesting.
Deliverables:
WCAG 2.2 AA Audit | Usability Test Plan | Test Synthesis Report
Outcome & Recommendations
This is a validated direction, not a finished product. Solid: the white space, the flow structure, the priority fix list. Not yet validated: the Access Needs segment (one participant only) and account-creation sequencing at the business-strategy level.
AI accelerated drafts, personas, journey scaffolds, simulated usability sessions, and synthesized requirement tables. I owned every decision that mattered: which questions to ask, what to trust, how to resolve conflicting persona needs, and labeling simulated findings as hypotheses, not proof.
Reflection
AI didn't just save me time generating — it sped up how fast I could kill weak ideas and expedite real strategic decisions. Simulated usability testing caught the personalization gap before a build sprint would have. The real value wasn't the AI's answers; it was the discipline of checking whether a "validated" claim actually was one.
The true balance came through applying human decision-making based on my more than 20 years of UX experience, coupled with creating targeted prompts to solicit rapid AI synthesis of research, design concepts, design, and validations that produced sound experiences that can evolve to production-ready design.
