CASE STUDY

Artemis

AI-Powered Strength Machine

Designing an AI-powered strength training system—balancing ambitious intelligence with real-world hardware, technical constraints, and shifting business priorities.

Designing an AI-powered strength training system—balancing ambitious intelligence with real-world hardware, technical constraints, and shifting business priorities.

My Core Ownership

  • Hands-on competitive research & synthesis

  • Voice interaction model for AI coaching

  • Assessment & goal-setting logic (collaborative lead)

  • End-to-end journey mapping (hardware + app)

  • Prototyping to validate interactions

Duration

6 months

Team

Steve Meadows (Principal Interaction Designer)
Marcus Mustafa (UX Director)
Chris Cannon (Principal UI Designer)
Louis Block (UI/Motion Design)

OVERVIEW

Quick Context

iFit set out to enter the connected strength training market—not by shipping another smart cable machine, but by building a system intelligent enough to understand users’ goals, constraints, and real-life disruptions.


I worked on UX research and strategy for Artemis, an AI-powered connected strength system designed to adapt to users’ goals, constraints, and real-life disruptions.

Over six months, I conducted hands-on competitive testing, designed end-to-end system architecture across hardware and app, and developed the assessment and voice interaction models.

The strategy, frameworks, and interactions I designed were approved by stakeholders and formed the foundation for future work.

6

Platforms analyzed through
hands-on testing

100+

Features mapped
against competitors

10

User journey maps
created

MARKET INSIGHT

Where existing platforms
fall short

Connected strength market was crowded. Real challenge wasn't adding another cable machine—it was creating genuinely intelligent experience.

What competitors consistently missed

Form tracking was superficial

Competitors flagged errors but didn't explain why or how to fix. Users wanted coaching, not correction.

Personalization stopped at onboarding

Platforms asked questions upfront but rarely adapted based on actual behavior, progress, or changing circumstances over time.

Goals were oversimplified

Fitness reduced to "build muscle" or "lose weight." Real users had complex goals encompassing mental well-being, recovery, and lifestyle integration.

Motivation faded quickly

Pre-recorded workouts felt impersonal. No platform made returning after breaks easy with personalized re-entry plans.

Mapping 100+ features across 6 platforms

Hands-on
competitive research

I led comprehensive competitive research, physically interacting with competitor machines to understand not just features but the actual user experience.

This hands-on approach revealed nuances that feature lists couldn't capture—like how Tonal's form corrections felt robotic, or how Tempo's camera positioning created awkward workout angles.

Testing Speediance firsthand to evaluate how form feedback, physical interaction design, and workout flow feel under real training conditions.

Evaluating form feedback and interaction flows on an internal prototype to surface feasibility and usability issues before production decisions.

DESIGN DECISIONS & CORE WORK

Rethinking assessment

Most connected strength platforms rely on “lift till failure” tests that measure strength in isolation. I designed a holistic assessment framework that evaluated the whole person—not just how much weight they could move in a single session.

A multi-dimensional assessment model

Physical capability

Strength baseline, range of motion, muscle imbalances, and movement quality.

Lifestyle context (captured via voice)

Injury history, current activity level, schedule constraints, and goals beyond physical appearance.

Continuous monitoring

Every workout contributes assessment data, supported by monthly “check-in” workouts that surface deeper insights without feeling like tests.

Trade-offs & Constraints

Trade-offs & Constraints

Assessment reimagined as an ongoing system, not a one-time test

Designing conversational AI

One of my primary contributions was developing the voice interaction model—defining how conversational AI should behave during workouts.

I mapped comprehensive scenarios across six categories: core workout commands, guidance & feedback, personalization, error handling, user fatigue & safety, and weight adjustments.

Trade-offs & Constraints

Trade-offs & Constraints

Balancing structured coaching with fast command execution

Command mode

Conversational mode

Mapping the complete experience

Artemis began as a standalone machine. Midway, iFit pivoted to a unified app ecosystem—breaking assumptions across onboarding, workouts, and progress tracking.

I created 10 end-to-end journeys spanning hardware + app, revealing gaps and dependencies before they became failures.

Journeys included: onboarding, assessment, workouts, progress, re-engagement, multi-user households, free lift, discovery, and more.

Architectural principals

Progressive disclosure

Flexible entry points

Persistent AI coach across touchpoints

Trade-offs & Constraints

Trade-offs & Constraints

Onboarding became the clearest window into how assessment, voice, and system constraints intersected.

IMPACT

What happened & what I learned

After six months, iFit decided to prioritize their unified app platform over Artemis hardware. We paused for several months, then resumed briefly before iFit moved the project to another agency.

What Got Delivered

Comprehensive competitive analysis, 10 complete user journey maps, information architecture, voice interaction model, assessment framework, scoring logic, UX strategy document.

What Worked

Hands-on competitive research provided strategic clarity. User journey maps bridged hardware and app experiences. Voice interaction model revealed user psychology. Constraint-driven design produced realistic solutions.

What I'd Do Differently

Test voice assumptions with users earlier. Advocate harder for phased feature rollout. Document decision rationale better. Prototype more around moments between workouts.