Rohto Pharmaceutical and Microsoft's innovation hub, "Microsoft AI Co-Innovation Lab Kobe," launch an AI project
– a co-creation initiative between customers and AI to commemorate The Mentholatum Company's 135th anniversary –
resulting in a heartwarming image created from over 1.5 million customer testimonials.
June 7, 2024
ROHTO Pharmaceutical Co., Ltd. (Osaka City, Osaka Prefecture, President: Masafumi Sugimoto) utilized Microsoft Corporation's (hereinafter referred to as Microsoft) innovation creation hub, "Microsoft AI Co-Innovation Lab Kobe" (Kobe City, Hyogo Prefecture, Director: Takehiro Hirai), to conduct a project by May 2024 to analyze and aggregate customer feedback using generative AI and create new value. We are pleased to report the results of this project.
The Mentholatum Company, a subsidiary of Rohto Pharmaceutical, leveraged its network of customers in over 110 countries worldwide to analyze and aggregate customer feedback received about the company and its products using AI. A scheme was then created to generate images from this anthropomorphic AI, and the final images were produced based on this data. The insights gained from this project will be used to proactively create new value by leveraging technology, such as applying the analysis of customer feedback to product and service development.
Background and details of this project
The Mentholatum Company, a subsidiary of Rohto Pharmaceutical Co., Ltd., was founded in 1889 and has since delivered its products to customers in over 110 countries worldwide. For many years, it has been a brand that understands and addresses the concerns of its customers. The company and Mentholatum products have received 1,525,571 comments from customers in 15 countries (United States, China, Canada, Hong Kong, Taiwan, Vietnam, Singapore, Malaysia, Thailand, Australia, United Kingdom, South Africa, South Korea, Nepal, and Japan).
This project utilized Microsoft's generative AI service, "Azure OpenAI Service," and based on the feedback we received, we designed and implemented the following scheme.
To achieve this, we used GPT-4 (large-scale language model), embedding-ada-002 (embedding model), and DALL-E (image generation model).
In the process of aggregating customer feedback, all customer feedback from each country was vectorized using an Embedding model and mapped onto a semantic space. Cluster classification was then performed using the k-means algorithm, and representative feedback for each classification was extracted using GPT-4. Furthermore, project members defined what aspects of customer feedback evoked emotion and gratitude, and based on these definitions, customer feedback was extracted using GPT-4. All customer feedback was aggregated by integrating each extracted feedback. In the process of personifying the AI, Mentholatum's character, the pure and compassionate Little Nurse, was defined as the persona that receives customer feedback, and prompt engineering was performed. First, the personified AI Little Nurse, who received the aggregated customer feedback, was made to generate a thank-you letter to the customers. Based on this, the personified AI Little Nurse was made to generate prompts for image generation, and images were generated using DALL-E (image generation model).
While customer feedback is usually categorized by product, this project involved extracting letters and images, allowing us to focus on customers' concerns and feelings. This process enabled us to analyze how customers were pleased with our products, even across different items. We believe that the AI's ability to interpret context and the use of embedding models will allow us to listen even more attentively to customer feedback and gain valuable insights for developing new products.
Similarly, we performed the same process for 15 countries based on customer feedback received in each country, and combined the generated images into a single image. Despite not being specified in the image generation prompt, an image filled with many hearts was generated.
As we celebrate Mentholatum's 135th anniversary, we reaffirm that our journey has been one shared with our customers. We will never forget our gratitude to our customers and will continue to dedicate ourselves to delivering products that will bring joy to customers around the world.
The cooperation of Microsoft AI Co-Innovation Lab Kobe was indispensable in carrying out this project.
The AI technologies, such as data analysis, obtained through this project will be used in future product development ideas and services to further enhance customer satisfaction.