BCG and Zeiss developed an application to help health care professionals supporting patients with information on elective treatments. Test user feedback has been strongly positive.
Zeiss Medical Technology has a long history of developing innovative devices, tools, and software solutions to advance the standards of care in a wide variety of medical specialties. Like professionals in most industries, many of the medical doctors and surgeons who are Zeiss’s customers struggle to find the time and resources to respond promptly and completely to inquiries from potential patients to ensure that these patients are adequately informed and educated about their options.
BCG and Zeiss recently teamed up to develop a proof-of-concept (PoC) generative AI (GenAI) application. The initial target: ophthalmologists and their clinical staff, who need help to respond to patient inquiries more easily and comprehensively. The goal was to create an application that will help increase patients’ interest in and understanding of available elective treatments such as modern laser vision correction procedures.
In just eight weeks, BCG helped Zeiss build a PoC GenAI application that generates responses to patient-submitted content forms and self-test forms. To ensure that the application’s responses are accurate and appropriate, the solution was designed to create answers based on a library of approved Zeiss marketing and patient materials. The application engages with patients in a personal, caring, and empathetic manner.
“When you’re dealing with medical-related questions about procedures or technology, it’s incredibly important to know that any information provided to patients is current, accurate, and pre-approved,” says Euan S. Thomson, head of the ophthalmology strategic business unit and the digital business unit for Zeiss Medical Technology. “Working with the expert team at BCG, we’ve been able to create a GenAI application based on our own approved materials. Those AI-generated responses are then validated, giving us great confidence in the quality of the application’s outputs and its utility for our customers and their patients.”
Test user feedback has been strongly positive. Practice managers have praised the responses produced by the GenAI application, saying they are likely to be valuable and helpful to their practices. One practice manager wrote, “While I initially thought the value of this would be limited to after-hours responders, I now feel like it would be great to use all of the time.”
In early and limited evaluations, 79% of responses produced by the GenAI application were assessed as being good enough to send directly to patients without any edits. When testers were given an opportunity to customize the application’s settings, the percentage of responses deemed patient-ready without needing any edits jumped to 93%.
Zeiss is further developing this project to achieve the company’s goal of helping patients feel more comfortable, informed, and empowered as they work with doctors to choose the best options to meet their unique health needs.
By giving doctors the ability to respond more rapidly and easily to patient inquiries, Zeiss’s GenAI application can help doctors and their office staff spend more time focusing on patient care, while potentially strengthening demand for future treatments such as vision correction procedures.
BCG is now helping Zeiss develop an overall GenAI roadmap and a new operating model so that the company can generate even more value from its deep AI capabilities.
BCG experts in artificial intelligence helped Zeiss assess a hundred possible GenAI use cases and identify the best option for a PoC build. We worked closely with Zeiss’s AI team to co-develop the GenAI algorithm and application using Azure OpenAI, Langchain, and chain-of-thought agent technologies with responsible AI guardrails in place to validate outputs. BCG used an agile project structure, incorporating customer insights and tester feedback to repeatedly refine the application for maximum impact.
Large language model (LLM) agent technologies
Large language models (LLMs)
Retrieval Augmented Generation (RAG)