Designing a Clinical-grade AI Companion: HealLink's Telehealth UX Case Study

AI · 6 min read

HealLink’s product team had to reconcile clinical safety with the efficiency gains of an AI companion that drafts visit notes and offers preliminary triage suggestions. Early prototypes that surfaced AI suggestions inline led to clinician friction: providers worried about liability and the time cost of verifying AI output. The team pivoted to a design that treated the AI as an assistive aide rather than an autopilot.

Key UX patterns included explicit provenance labels on any AI-generated content, a two-step verification workflow where clinicians could accept, edit, or reject suggestions, and an audit trail that recorded AI contributions for later review. Patient-facing explanations were simplified: the app explains when AI influences triage and offers clinicians a transparent summary to share. HealLink also implemented confidence scoring and a ‘flag for second opinion’ action for low-confidence recommendations.

Clinical trials across three partner clinics showed that average visit documentation time fell by 28% and clinician trust scores (measured via surveys) improved by 12 points when the human-in-loop controls were present. Importantly, no critical errors were traced to AI suggestions during the pilot, validating that UX patterns—visibility, control, and straightforward error recovery—are essential when designing AI features in regulated, high-stakes domains.