Specialist healthcare provider: Building a digital twin for patient education

1 min read

Doctor using a laptop during a patient consultation

A specialist healthcare provider wanted patients to arrive at consultations better informed about breast reconstruction options without adding pressure to busy clinical staff. The aim was to make trusted information easier to access before in-person visits, while keeping the experience safe, empathetic and clinically grounded.

A web-embedded AI experience was designed to combine a multilingual chat assistant with a voice-interactive digital twin of the lead surgeon. Powered by IBM WatsonX and HeyGen, the assistant gave patients a more natural way to ask questions, understand options and prepare for conversations with the clinic.

Outcome 1: Delivered MVP roadmap planning in eight weeks, taking the experience from end-to-end design to deployment planning
Outcome 2: Supported three languages across chat and voice, giving patients more accessible ways to interact
Outcome 3: Orchestrated two large language models to support safe, empathetic and surgeon-like responses

The challenge

Patients considering breast reconstruction often need clear, sensitive information before they feel ready for a consultation. But answering common questions manually can add pressure to clinical teams already managing busy schedules.

Our client needed a patient education experience that could provide consistent guidance without collecting personal health data or replacing clinical care. It also needed to feel natural, handle multiple languages and give patients a clear path back to the clinic when human support was required.

The approach

The programme began by defining the core patient journey, from seeing an educational disclaimer to choosing Chat or Voice within a single multilingual interface. This helped set expectations clearly while giving patients control over how they wanted to interact.

A knowledge base was then built from approved clinic brochures, FAQs and documents, creating a single source of truth for the assistant. This ensured responses stayed grounded in curated practice materials rather than unsupported information.

Our team implemented two large language models: one to manage orchestration and guardrails, and another to respond in the surgeon’s tone. This allowed the assistant to keep answers accurate and compassionate while maintaining the clinical boundaries needed for patient education.

The experience also integrated speech-to-text, text-to-speech and a HeyGen video avatar, enabling patients to speak to the digital twin and receive video replies. IBM WatsonX supported the underlying AI experience.

Guardrails, testing and review steps were designed so clinicians and compliance teams could validate responses, tune content and safely evolve the assistant over time.

The impact

Patients gained 24/7 access to trusted education before in-person visits, with options to ask questions through text chat or spoken conversation. This made the experience more accessible and helped patients engage with breast reconstruction information in the format that felt most natural to them.

The assistant also reduced pressure on clinical staff by handling common education-led questions while keeping answers grounded in approved materials. HIPAA alignment was maintained by avoiding the collection of personal health data, with a clear path for patients to contact the clinic when needed.

The work established a scalable, modular architecture and monitoring framework that can support future clinical use cases. Ultimately, our client was better positioned to provide accessible patient education, protect clinical trust and safely extend AI-enabled support over time.