Haleon (LSE / NYSE: HLN) is a global leader in consumer health, with a purpose to deliver better everyday health with humanity. Haleon’s product portfolio spans five major categories - Oral Health, Pain Relief, Respiratory Health, Digestive Health and Other, and Vitamins, Minerals and Supplements (VMS). Its long-standing brands - such as Advil, Sensodyne, Panadol, Voltaren, Theraflu, Otrivin, Polident, parodontax and Centrum - are built on trusted science, innovation and deep human understanding.

The healthcare landscape is evolving, with a growing emphasis on personalized consumer experiences. Haleon, built on the foundation of scientific expertise, innovation, and deep human understanding, aims to harmonize consumer insights with cutting-edge healthcare service innovation. To ensure a premium customer experience for its vast consumer base, the company sought to implement a fully automated, real-time service delivery system. This technologically-rich solution would address each consumer's enquiry and demand instantly, eliminating delays and enhancing overall satisfaction.

Haleon aimed to develop an intelligent, responsible, and professional system that could handle a large volume of consumer enquiries in real time, ensuring a premium customer experience without compromising personalization or efficiency.

With healthcare at the forefront of the AI revolution, the nutrition, medicine development, and health management industries have largely begun leveraging Large Language Models (LLMs) to unlock unprecedented value.

By harnessing Alibaba Cloud's sophisticated LLM, Tongyi Qianwen (Qwen), and integrating it with Haleon's proprietary knowledge graph through retrieval-augmented generation, Haleon’s solution now seamlessly integrates across all customer touchpoints, ensuring a cohesive end-to-end journey and hyper-personalized healthcare management.

Leveraging the Qwen large language model, integrated with Haleon’s existing knowledge graph, Alibaba Cloud supported Haleon to build AI nutrition assistant services through retrieval-augmented generation. The solution includes end-to-end design, development, and integration across the stakeholder spectrum and implements the following components:

Qwen Large Language Model and Prompt Engineering:
Through Prompt Engineering, the input and output of the Qwen model are arranged and connected to realize question pre-processing, personalized Q&A based on user tags, structured nutritionist responses, specified tone styles, and other features.

Vector Recall:
Sort out and split the data from the enterprise's exclusive knowledge base, and vectorize the knowledge through the Embedding model. The purpose is to recall unstructured knowledge based on user questions and use it to enhance the answers of the large language model.

Knowledge Graph Recall:
Knowledge graph recall makes full use of the existing data assets of the enterprise and uses NL2Cypher to retrieve knowledge such as relationships and attributes that are beneficial to question answering from the knowledge graph. This improves the professionalism and reliability of the large language model's answers.

Recall Sorting:
Integrate multi-channel recall results (including vector recall, sub-graph recall) to ensure the reliability of the retrieval results while eliminating any noise, thereby enhancing the answering effect of the large base model.

Front-end Integration:
By integrating Qwen's standard API interface with the front-end system, multi-touch service integration is achieved, particularly with Enterprise WeChat. This integration provides functions including login, authorization and authentication, messaging, and security encryption, among others.

The partnership between Alibaba Cloud and Haleon, powered by Tongyi Qwen, has revolutionized personalized healthcare for millions. This collaboration enabled Haleon to achieve unprecedented precision and insight in addressing health inquiries

Nutritionists utilizing the AI nutritional assistant experienced a substantial enhancement in professionalism and scientific rigor, alongside notable gains in service efficiency. Currently, a single nutritionist supported by this technology can serve over 1,300 consumers, representing a six-fold increase in efficiency compared to traditional models where human nutritionists typically handle around 200 clients each on other platforms. Additionally, they were able to respond to inquiries in the user’s preferred language.

Haleon plans to further enhance its AI-powered health management services by expanding its capabilities of knowledge graph x LLM and improving personalization. Leveraging Alibaba Cloud's scalable computing resources and proprietary AI frameworks, Haleon aims to accelerate innovation in consumer healthcare, emphasizing data-driven insights and intelligent automation throughout its value chain.

In addition to the AI-powered nutritional assistant project already underway, Haleon also looks forward to collaborations involving AI-generated content, AI brand ambassadors, AI-personalized experiences, and exploring how to apply large language model technologies in conjunction with Haleon's trusted science to empower healthcare professionals.