Shift and Mining Dining Hall Optimization with AI

Imagine a mining operation in northern Chile where 1,200 workers finish their shift at 1:00 p.m. They are all hungry, they all want to eat quickly and they all head to the same place: the site dining hall. The result is predictable and repeats day after day, shift after shift. Lines of 30 to 40 minutes, frustrated workers who lose part of their rest time waiting, an overwhelmed kitchen team trying to serve hundreds of plates simultaneously, and a significant amount of food that ends up in the trash because actual demand does not match the preparation projections. This problem, which seems minor compared to the technical challenges of mining extraction, has a direct impact on productivity, staff satisfaction and the site’s operating costs.

The Problem: Systematic Congestion at Peak Times

At mining sites that operate on 7x7, 14x14 or 4x3 shift schedules, meal times create unavoidable bottlenecks. The dining hall has a limited physical capacity, typically between 250 and 350 seats, but it must serve more than 1,200 people in very narrow time windows. The shift change concentrates all the workers within an interval of just 90 minutes, and without an intelligent management system, the distribution of arrivals follows a bell-shaped pattern: a few arrive at the start, most concentrate in the central 30 minutes, and a few arrive at the end when little variety of food is left.

The consequences go beyond discomfort. A worker who spends 40 minutes in line has 40 fewer minutes of real rest, which affects their recovery and their performance in the next shift. For the kitchen team, demand peaks generate stress and preparation errors. And for the company, food waste represents a direct cost that in large operations can exceed 200,000 dollars per year. In addition, staff satisfaction surveys consistently identify the dining hall experience as one of the main sources of dissatisfaction, which impacts talent retention in an already competitive labor market.

The Solution: A Scheduling Agent with Mobile Notifications

The solution we implemented at ValueData consists of an AI agent that autonomously manages the distribution of meal times. The agent operates in three complementary layers that, combined, completely transform the mining dining hall experience.

The first layer is predictive planning. The agent analyzes historical dining hall attendance data cross-referenced with shift planning, the site calendar, weather conditions and special events such as training sessions or visits. With this information, it generates an accurate projection of how many people will attend each service and automatically distributes the workers into 15-minute time windows. The assignment is not random: it respects the operational constraints of each area, the travel times from the different points of the site and the preferences registered by the workers themselves.

The second layer is real-time communication. Each worker receives a push notification on their mobile phone or site device indicating their assigned time window for each meal. The notification includes the recommended time, an estimate of the current waiting time and alternatives if the worker cannot attend within their assigned window. The system is flexible: if a worker needs to change their time for operational reasons, they can request a change from the app and the agent reassigns in real time while maintaining the load balance.

Monitoring with Computer Vision

The third layer, and perhaps the most innovative, is real-time monitoring through computer vision. Cameras installed at the entrances and inside the dining hall feed a detection model that counts the people in line and estimates table occupancy. This information is processed every 30 seconds and feeds the scheduling agent, which can make dynamic decisions such as bringing forward or delaying the notifications to the next group, activating an additional service line if kitchen staff are available, or sending an alert to the operations team if the situation moves outside normal parameters.

The computer vision model does not identify individual people nor store images of faces, complying with privacy regulations. It only counts people and analyzes flow patterns: how many people enter per minute, how long on average they remain in line, and what the table turnover rate is. This data, combined with the agent’s history, enables continuous learning that improves the predictions week after week.

Measurable Results

The results obtained in the pilot implementation speak for themselves. The average waiting time in the dining hall was reduced from 38 minutes to 8 minutes, a 79% improvement. Food waste decreased from 24% to 9% of the total prepared, thanks to more accurate demand projections that allow the kitchen team to adjust preparation quantities. Staff satisfaction related to the food service rose from 45% to 82% in the monthly surveys. And a particularly relevant figure for management: workers’ effective rest time increased by an average of 25 minutes per shift, which translates into better well-being indices and, according to the safety metrics, a reduction in minor incidents associated with fatigue.

From a financial standpoint, the reduction in food waste generated direct savings of approximately 180,000 dollars per year. But the most significant impact is intangible: better-rested workers, less staff turnover and an operation that demonstrates that technology can improve daily life at the site, not just the productivity of the heavy equipment.

Beyond the Dining Hall: A Replicable Model

The most valuable thing about this approach is that the same scheduling agent model can be applied to any shared resource within a mining site: transport buses, showers, laundries, gyms and training rooms. The underlying logic is the same: predict demand, distribute access intelligently, communicate with users in real time and adjust dynamically based on the current situation. At ValueData we believe that digital transformation in mining does not begin with the large industrial automation systems, but with the solutions that directly impact the experience of the people who work at the site every day.