How I convinced skeptical medical teams that AI could enhance, not replace human care.

Over six months, I led the redesign of the app, strategically integrating AI to personalize the experience while ensuring that the new interface remained familiar and intuitive, improving user confidence in remote rehabilitation.

Timeline

April 2022 - October 2023

Role

Lead UX Designer

Method

Atomic Research Model for complex stakeholder ecosystems

Team

14 members across design, development, QA, project management, and client teams

Business problem

A rigid, impersonal experience that failed to adapt to the individual needs of patients led to low engagement and decreased confidence in remote rehabilitation. Limited communication between patients and clinics created friction, hindering adherence to rehabilitation plans, negatively affecting patient outcomes, and consequently, retention.

What We Got Wrong

Failed hypothesis #1

Video exercise demonstrations would be sufficient to perform movements correctly.

Failed hypothesis #2

Automated reminders and notifications would improve appointment attendance.

Failed hypothesis #3

One-size-fits-all AI personalization.

Controversial Decisions

The Build vs. Band-Aid Choice

I defended 4 months additional development and new technical infrastructure against leadership's preference for quick wins.

Why this was controversial

Required convincing skeptical medical teams, additional budget, and regulatory compliance considerations when competitors were launching faster solutions.

The AI Ethics Stand

Called for clinical oversight for all AI exercise modifications instead of fully automated adjustments, maintaining human judgment in medical decisions.

Why this was controversial

AI development was delayed, operational costs grew due to clinical reviews, and tech teams pushed back, favoring scalable automation over manual oversight.

Apr

Medical team feared AI would replace them → I designed research to prove AI enhances clinical expertise.

May

Stakeholders wanted to skip testing for faster launch → I defended user validation despite external budget pressure.

Jun

Stakeholders resisted design changes → I used data from 41 users to justify interface redesign.

Jul

Developers wanted to build basic notification system → I advocated for conversational interface despite delay.

Aug

AI couldn't distinguish between different injury types → I created user-input overlays to contextualize recommendations.

Approach

The Atomic Research Method

Traditional research often gets buried in reports or distorted by bias, breaking the link between user feedback and design decisions. This approach helped build trust with stakeholders skeptical of both AI and design.

No More Missed Appointments

We designed an AI-powered flow that let patients confirm, reschedule, or cancel appointments via an AI assistant, reducing no-shows and giving them greater ownership of their recovery.

Move, Measure, Improve

We used an AI motion-tracking assistant to deliver real-time feedback and personalized session summaries, keeping patients informed, motivated, and progressing.

15%

Increase Patient Retention

Increase patient retention by 15% over the next 12 months, ensuring higher adherence to rehabilitation.

10%

Reduce Missed Appointments

Achieve a 10% reduction in missed appointments within the first six months of product use.

Improve Patient Outcomes

Improve patient outcomes and consequently increase the credibility of the platform.

Where We’re Going

Metrics → Meaning

Provide clinics with a real-time dashboard displaying patient risk trends, enabling more efficient, targeted outreach.

Inclusion → Impact

Incorporate users with different levels of digital literacy into the research to develop a more accessible and inclusive design.

Engagement → Adherence:

Introduce game-like rewards, streaks, and personalized achievements to motivate patients, transforming repetitive exercises into rewarding milestones.