Technology
Building an AI assistant that doctors trust enough to let it handle their patients' questions — and that patients trust enough to use.
Built an AI healthcare assistant for doctors and patients in Hong Kong, where 82% of questions were resolved automatically through the knowledge management system (KMS) without needing direct doctor involvement.
headquarter
how we support
Project Duration
Overview
Building an AI assistant that doctors trust enough to let it handle their patients' questions — and that patients trust enough to use.
Dr Baba is an AI-powered healthcare assistant built for the Hong Kong market — designed to handle the high volume of routine patient questions that consume clinical time without adding diagnostic or therapeutic value. The challenge was trust on two sides: doctors needed confidence that the system would handle routine questions accurately and escalate appropriately, while patients needed assurance they were receiving reliable health information. Noran Design approached this as a knowledge design challenge as much as a product design challenge — building the knowledge management system that underpins every response. The outcome was a continuously improving platform that satisfied clinical governance requirements while delivering measurable time savings for doctors and clear, accessible support for patients.
Challenge
Building an AI that doctors trust with their patients.
The knowledge management system (KMS) was the critical component — the mechanism by which the AI's responses were grounded in clinically validated information and continuously improved based on real-world usage. Doctors needed to trust that the system would handle routine questions accurately and escalate appropriately. Patients needed to trust that they were receiving reliable health information and not being dismissed by a chatbot. Getting both right simultaneously was the core design challenge.
Solution
Build the knowledge system first. The interface is the easy part.
We designed the knowledge architecture underpinning the AI's responses — structuring medical information to enable accurate, contextually appropriate answers across the full range of routine patient questions. Alongside this, we built the escalation framework: the decision logic determining when a question required human clinical judgment and how that handoff was handled clearly and safely. The patient-facing conversational interface was designed for clarity and trust — warm and accessible without overstepping the system's validated clinical scope. The full system was then validated with clinical users, iterating until the 82% automatic resolution rate was achieved with confidence.
Impact
82% of patient questions resolved automatically. Zero clinical safety incidents.
82% of routine patient questions resolved automatically through the knowledge management system — without direct doctor involvement. Significant reduction in non-clinical doctor workload, returning clinical time to patients who genuinely needed it. A trust architecture that satisfied both the clinical governance requirements of the medical practice and patient expectations for reliable health information. A continuously improving system, designed to learn from real-world usage and improve response accuracy over time.
"What Noran built for us wasn't just a chatbot — it was a clinical knowledge system that we trust to represent us to our patients. The 82% automatic resolution rate speaks for itself, but what matters more is that we've had zero incidents we wouldn't stand behind."
Dr Baba
Clinical Director, Dr Baba
Dr Baba — AI Healthcare Assistant