Developed an “AI Nutritionist” Feature for Company D’s Diet Coaching App

Healthcare Platform Company D, 
“We want to develop an AI nutritionist that assists coaching consultants in suggesting healthy meal plans to clients and responds promptly and accurately to their questions!”

Client D is,

A beauty and healthcare platform company that develops a diet coaching app through joint research with a leading diet-focused Oriental medicine clinic.

Project Overview 

Company D’s app guides users through carbohydrate-restricted dieting, offering comprehensive weight tracking, nutritional analysis, and 1:1 personalized coaching. Company D aimed to enhance the app by incorporating generative AI, adding features like automatic meal plan analysis, healthy meal substitutions, and more effective 1:1 coaching services. Before seeking out AI-specialized companies, Company D had established specific objectives, and among the various options considered, it ultimately chose Crowdworks.

The Objectives:

• Create a personalized AI service for customers to enhance satisfaction: We aim to provide faster, real-time access to the services and information that customers need, and we aim to secure competitiveness by developing services that differentiate us from others.
• Improve the efficiency of our consultants and staff with AI: We aim to reduce the repetitive tasks of our consultants who assist many customers daily, allowing them to focus more on specialized tasks by leveraging generative AI.
• Leverage data-driven insights for business development: We aim to systematically analyze the data accumulated through customer interactions and services to provide personalized, high-quality services, as well as use the insights for trend analysis and service improvements.

Role of Crowdworks

• Crowdworks played a pivotal role in consulting and project design, handling everything from requirements definition to system architecture planning, scenario design, and service functionality mapping.
• We analyzed existing data of Company D to prepare all the necessary data for service development.
• We utilized LLMs to build customized AI services for the company. 
• We tested the model’s performance, collected and incorporated both qualitative and quantitative feedback from the client, and continuously improved its functionality.

From analyzing and evaluating customers’ meal plans to providing useful answers, our AI service covers it all!

1) ‘AI Diet Advisor’ Optimized for Analyzing, Evaluating, and Suggesting Diets

This feature allows customers to upload their meals to the app via ‘photos’ or ‘text.’ AI then automatically analyzes and evaluates the nutritional content, providing feedback and recommending a personalized diet plan based on the customer’s gender, age, BMI, and other factors. Through this process, customers receive accurate and detailed diet analysis and can receive real-time recommendations for new, healthy diet options tailored to their personal needs.

► The key component to developing this service was ‘data.’ We gathered and reprocessed various types of data, including dietary information from the largest diet-focused Oriental medicine clinic in the country, nutritional evaluation details, alternative healthy diet options, and personalized comments based on individual data. Additionally, we designed answer scenarios based on this data and wrote corresponding prompts. To refine the service, we worked with multiple consultants and relevant staff members, thoroughly understanding their workflows and data, and made multiple revisions based on feedback after development.

2)  ‘AI Chatbot’ That Swiftly Answers All Customer Questions

The AI chatbot was also developed. Designed to quickly resolve any customer inquiries about dieting, when asked questions related to topics regarding membership details, surveys, health status, diet, and exercise, it provides accurate responses. The development of this service incorporated *NL2SQL and persona prompt techniques, enabling the AI model to refer only to the necessary information in the customer database and generate relevant answers. If a question asked is unrelated to the preset topics, the chatbot will respond with the pre-configured answers set by Company D.

*NL2SQL: A technique that converts natural language into a database query language, allowing the database to understand everyday language-based questions.

“I love how it suggests nutritious alternative diet plans, not just the same old meals.” – Customer A

“With fewer repetitive tasks, I can focus on delivering higher-quality, professional coaching!” – Consultant B

Drawing from its extensive experience in enterprise AI development across various industries, Crowdworks minimized the persistent hallucination issues of LLMs, enhancing the quality and accuracy of the answers through performance validation and results data, leading to improved customer satisfaction. With the new service, consultants at Company D have been able to provide more structured and effective coaching, and though the service is still in its early stages, employee utilization and satisfaction are already high. In the case of the AI Diet Advisor, it achieved a high accuracy rate of over 92% in Company D’s internal evaluation. Thanks to this, Crowdworks continues to collaborate with Company D to further enhance the AI service.

Explore how generative AI can transform your healthcare services. Contact Crowdworks today to consult with our AI specialists!