Akki — AI Health Assistant (Syni.ai)
Where clinical precision meets empathetic AI for chronic care management.
2x
Engagement Uptake
90%
Automation Rate
100%
Red-Flag Success
75%
Time Reduction
Hero & Overview
Akki is a WhatsApp-based AI assistant designed to bridge the gap between clinic visits for patients with hypertension, diabetes, and asthma. I led the end-to-end product and conversation design, ensuring the assistant provides professional-grade health coaching while remaining accessible and culturally resonant.
The Multi-Stakeholder Research Framework
The design logic for Akki was built on a dual-track research process: deep-dive user testing and high-level expert consultation.
Track A: User-Centric Testing (The 'Why')
- Mothers (Pregnancy & Early Motherhood): Researched the cognitive load of the motherhood journey. We found that 70% of mothers were overwhelmed by manual health tracking. This informed our 'Nurture Flow'—automating prenatal vitamin nudges and symptom-relief advice.
- Diabetes & Hypertension Patients: Identified a 40% drop-off in medication adherence during 'routine' weeks. We solved this with 'Incentive-Based Nudges.'
- Caregivers: Addressed the 'silent symptom' anxiety (reported by 85% of caregivers) by designing automated red-flag triggers.
Track B: Expert Consultation (The 'How')
- Clinical Doctors: Worked closely with physicians to map out Red-Flag Escalation Logic. This ensured that the AI only handles routine queries while immediately flagging high-risk vitals to a clinician's dashboard.
- Nutritional Experts: Collaborated to build a Cultural Nutrition Library. We moved away from generic 'Western' diets to map out localized Kenyan meals (e.g., Sukuma Wiki, Managu, Ugali) with accurate glycemic and caloric data.
Data-Driven Impact
2x Engagement Uptake
By implementing localized meal suggestions from our nutritional experts, we saw double the engagement compared to generic health advice.
Clinical Safety Net
100% of 'red-flag' symptoms identified in testing were successfully escalated to the clinician dashboard within seconds.
90% Automation Rate
The conversation design handles 9 out of 10 queries autonomously, leaving doctors to focus only on high-risk interventions.
Friction Reduction
Shifted from open-text logging to 'Quick Reply' buttons, reducing user logging time from 60 seconds to under 15 seconds.
Key Deliverables
Clinician Dashboard Suite
Designed widgets for risk-priority flags, adherence scores, and longitudinal BMI sparklines.
Empathetic Tone Guide
A framework for 'Localized Nudges' that uses inclusive, non-judgmental language.
Protocol-Driven Conversation Flows
Mapped out safe, clinical-approved paths for medication logging and symptom reporting.
What I Learned
- Localization is Personal: Users don't just want health advice; they want advice that fits their kitchen and their culture.
- Expert-Led Design: Co-designing with doctors isn't just about safety—it's about ensuring the digital tool fits the real-world capacity of a busy clinic.
- Conversation Design Matters: The difference between 60-second logging and 15-second Quick Replies isn't just efficiency—it's the difference between consistent use and abandonment.
- Red-Flag Logic Saves Lives: Building clinical escalation isn't a feature—it's the core safety mechanism that makes AI in healthcare possible.
Project Gallery
