5.4 Artificial intelligence in analysis

Artificial intelligence models in restaurant and foodservice operations generally rely either on supervised machine learning or unsupervised machine learning. In supervised learning, historical data such as previous sales, weekdays, seasonality, weather conditions, and special events are used to train models that predict future demand. Examples include Random Forest, linear regression models, and LSTM neural networks, all of which have been used successfully to forecast sales volumes in restaurants and catering settings.

In unsupervised learning, the goal is not to predict a predefined target variable, but rather to identify hidden patterns and relationships in the data. These methods can be used to detect links between food waste, production routines, ordering patterns, product combinations, and operational workflows. For example, they may reveal recurring deviations associated with shifts, weekdays, or product groups that are linked to higher waste levels or systematic overproduction.

Research indicates that improved demand forecasting can significantly reduce food waste. In a recent study of catering and foodservice operations, more accurate forecasting models, including Random Forest and LSTM, were associated with a potential waste reduction of 14–52% compared with older and less precise planning methods. Studies in restaurant settings have also shown that machine learning can provide practical and competitive sales forecasts for both short-term and week-ahead planning.

In practice, these models often appear not as standalone academic methods, but as part of operational software platforms. Systems such as Tenzo use machine learning to forecast demand based on sales, weather, holidays, and local events, while Apicbase connects those forecasts directly to inventory levels, recipe data, and automated purchasing workflows. Solutions such as Orbisk complement this by using image recognition and waste analytics to identify actual food waste patterns, helping managers uncover hidden links between waste and kitchen workflows

Tenzo: What it does: Forecasts demand and sales based on historical data, weather, holidays, and events. It is useful for improving staffing, purchasing, and prep planning. : Link: https://www.gotenzo.com/

Apicbase: What it does: Connects demand forecasting, inventory management, recipes, and purchasing. It is well suited to estimating quantities more accurately and reducing overproduction and waste. : Link: https://get.apicbase.com/demand-forecasting-restaurant/

Orbisk: What it does: Analyses food waste through image recognition and data analysis. It is useful for identifying patterns in what is being discarded, when it happens, and potentially why. : Link: https://orbisk.com/blog/food-waste-management-system-for-hospitality/