Planeat.eco

CHALLENGES

The goal of Planeat.eco was twofold: on the one hand, to increase customer retention and loyalty while reducing friction; on the other hand, to minimize the risk of loss in earnings. Specifically:

• Eliminate the risk of users forgetting to place lunch orders in time, especially on Fridays for Mondays’ meals.
• Reduce the drop rate caused by the time required to browse the entire menu and select dishes.
• Enhance the user experience when choosing and planning meals.

SOLUTION

beSharp has developed a next-generation daily and/or weekly planner based on Generative AI using Amazon Bedrock. LLMs have paved the way for a completely new user experience that is far better than classical recommendation engines’ experience based on NLP algorithms, best fit, and recommendation. The solution includes:

• Possibility for users to use natural language to express their food preference criteria.
• Independence from labeling, collaborative filtering, and large amounts of historical data.
• Implementation of a meta-prompt that can provide the system with details that are essential for the solution to work, including the entire menu available on the platform. This part remains invisible to the users, ensuring the perfect balance between UX and cost control.

OUTCOMES

• Natural language
  Highest level of customization and accessible UX
• 100% fitting results
  User expectations are fulfilled thanks to the criteria provided through the prompt
• Time-saving
  The solution helps save time by reducing decision-making time. This increases the conversion rate.
• Minimal implementation effort
  Through LLM, tasks such as model creation, selection, and training can be bypassed. This minimizes time-to-market and offers a real competitive advantage.
• Maximization of company earnings
  The planner ensures daily earnings, which are always optimized since the model plans the meals using the maximum number of credits possible, according to the 80 credits/day available.
• Zero waste
  Thanks to targeted recommendations, no dish is wasted.

MAJOR KEY PERFORMANCE INDICATORS (KPIs)

• Order time reduced from 15 minutes to 2 minutes.
• User spending is reduced by up to 45% compared to having lunch in a restaurant or ordering from other services based on the same food.

NEXT STEPS

• Accessibility
  Planeat.eco has always been committed to accessibility. Therefore, a vocal prompt will be available soon.
• Wider service availability
  By the end of the year, the mobile app feature will be released, and the Gen-AI-based service will also be available for families.
• Continuous Improvement
  The company aims to improve output quality through techniques like RAG (Retrieval-Augmented Generation) that allow LLMs to leverage additional data without the need for model retraining.