Leveraging AI to improve operations for a leading restaurant chain

Founded in 2016, iLunch is a Lithuanian smart restaurant chain operating around 45 locations across Lithuania, Latvia, Poland, and the UAE. The company specializes in a high-efficiency lunch model that uses self-service technology and a daily rotating menu to serve freshly prepared hot meals in under two minutes, primarily catering to the fast-paced needs of business centers and urban professionals.

The restaurant group operates in a demand-driven environment, which often leads to inefficiencies and difficulties in operational planning. Maintaining consistent dish quality across branches proved to be challenging, which impacts customer’s experience and brand image. While iLunch had extensive database of sales, operational, and customer-related data due to underdeveloped analytical tools, they struggled to extract actionable insights. The client partnered with Civitta to develop a suite of AI-driven tools tailored to address these issues.

Building on the BI data analytics infrastructure our team developed in the first iLunch project, we implemented even more advanced data features to elevate the tools’ analytical capabilities. The resulting AI, machine learning, and predictive forecasting features empower management to conduct deeper real-time analyses of their business.

• Phase I: Restaurant-Level Demand Forecasting

We developed an advanced AI model to predict daily customer footfall at each branch. The model integrated:

– Internal data: historical sales, receipts, pricing, and quantity of items sold

– External data: weather conditions, holidays, weekends, special occasions (e.g., Christmas), and health factors (e.g., flu season)

After thorough data cleaning and preprocessing, individual forecasting models were built for each restaurant using machine learning techniques. These models enabled the client to optimize staffing, inventory planning, and procurement strategies, reducing waste and improving service efficiency.

• Phase II: Menu Item Purchase Prediction

Building upon customer traffic predictions, we developed a second model to forecast what items customers were likely to order. The model considered:

– Seasonal preferences and weather patterns

– Active promotions and special deals

– Historical purchasing behavior by location

This enabled the client to better plan kitchen operations, align marketing efforts with expected demand, and reduce stockouts and overproduction.

• Phase III: Dish Quality Analysis with Machine Learning

To ensure consistent quality across branches, we integrated a tool that analyzed dish photos captured at the point of preparation. The system was trained to:

– Verify that the dish matched the customer’s order

– Assess presentation quality against predefined standards

The tool provided feedback to kitchen staff and aggregated quality compliance data for management reporting and training initiatives. The tool was developed by a third party with whom we cooperated in this project. 

• Phase IV: Customer Experience Insight Using Generative AI

We employed Generative AI and natural language processing (NLP) models to transform unstructured customer comments into actionable insights. This system:

– Collected reviews from digital platforms and feedback forms

– Classified comments into key categories (e.g., service, food, ambiance)

– Analyzed sentiment and tone to assign satisfaction scores

These insights enabled the client to prioritize improvements, address recurring issues, and enhance customer loyalty initiatives.

Civitta’s work has improved client’s operational efficiency by optimizing staffing and inventory, which reduced food waste and improved cost control. By predicting which menu items customers were most likely to order, the client was able to avoid stockouts of popular dishes and run kitchen operations more efficiently. The dish quality analysis tool introduced a standardized quality control framework across all branches, strengthening customer trust. In addition, Generative AI transformed unstructured customer feedback into clear, actionable insights, supporting continuous improvement of the overall customer experience.

The case demonstrates that AI initiatives deliver greater value when designed as a unified approach rather than as separate tools. Integrating demand forecasting, quality assessment, and customer feedback analysis created a more consistent and aligned foundation for operational decision-making.

By focusing on branch-level, Civitta ensured that AI models were not only technically robust, but practical and relevant for operational teams. This integrated approach is well suited to other demand-driven industries with complex operations and large volumes of data, where AI can support better coordination, consistency, and informed decision-making.