Shift Planning with Artificial Intelligence: How to Build the Right Team, Not Fewer Staff in HORECA?

It’s easy to know that the restaurant will be full on Saturday night. The harder part is accurately predicting how many servers, bartenders, and kitchen staff will be needed between 6:00 PM and 7:30 PM.

Service can falter when there is one staff member short. When more staff than needed are scheduled, labor costs rise. The fundamental issue in shift management for HORECA businesses has been maintaining this balance for years.

The guide published by workforce management software provider Deputy in August 2026 defines AI not as an automatic decision-making system in this equation, but rather as a helper that forecasts demand and prepares shift suggestions for managerial evaluation. Deputy states that historical trading and demand data can be used to predict staffing needs.

This approach can be meaningful for restaurants, cafes, bars, and hotels. However, technology is only one part of a good shift plan.

First, demand needs to be forecasted

Traditional shift planning in most businesses relies on past experience.

The approach of "We were busy last Friday, let’s schedule the same team this Friday" is still common in small businesses. However, actual demand can vary based on weather, number of reservations, events, season, promotions, and time of day.

The core promise of AI-supported workforce management systems is to create a more detailed staffing needs forecast by combining some of these variables with historical operational data.

According to Deputy’s own statement, systems can create draft shifts from data such as demand, employee availability, and labor costs. This information is based on the company’s own product approach; it does not automatically guarantee lower costs or a better employee experience.

The independent HR organization SHRM also states that AI-supported workforce planning can be used in areas such as smart shifts, evaluating employee preferences alongside labor costs, and automating high-volume administrative tasks.

A good shift is not one set with the least number of employees

One of the most dangerous goals in planning with AI is to use the system solely as a tool for "reducing labor hours".

A restaurant operating with eight staff instead of ten may reduce personnel costs on paper. However, if table turnover time increases, order errors rise, or employees are constantly working at capacity, overall operational performance can deteriorate.

Therefore, in workforce planning, not only labor cost percentage but also service performance should be monitored.

For example:

• Sales per hour

• Sales per employee

• Table turnover time

• Order preparation time

• Overtime

• Last-minute shift changes

• Employee turnover

• Guest complaints

When evaluated in the same table, it becomes clearer whether low labor costs reflect real efficiency or a lack of staffing.

SHRM's workforce planning studies also emphasize that AI and automation can support human teams, but managerial oversight remains critical.

Skill planning may be more important than the number of people

Having six people on a shift does not mean the right six people are present.

Especially in restaurants and hotels, employees do not have the same skills.

While a bartender can manage a busy cocktail service, a new team member may need support at this station. The capacity of an experienced server in banquet operations cannot be equated with that of a worker on their first shift.

Therefore, in advanced workforce planning, alongside the number of people, the skill matrix becomes important.

What competencies are needed at a specific hour of the shift?

How many people can open the cash register?

Who knows the closing procedures?

Which employee is trained in allergen procedures?

Who can work at the bar or coffee station?

AI-supported systems may account for these variables more in the future; however, if skill data is not up-to-date, the algorithm's suggestions may also be incorrect.

Technology cannot turn poor human resource data into a good plan.

Employee preferences should be part of planning

Another dimension of shift optimization is employee experience.

A system can create the mathematically ideal schedule for the business; however, if it continuously neglects employees’ availability, need for rest, or preferences, it may not be sustainable in practice.

SHRM states that mobile workforce technologies can support shift management, training, and employee communication for frontline workers; flexible shift options are also part of the employee experience.

Therefore, the goal of AI should not only be to find out "which employee is available".

Businesses should also consider the following information if possible:

preferred working hours, maximum weekly hours, consecutive shift load, leave requests, task competencies, and frequency of last-minute changes.

Especially for businesses in HORECA facing employee retention issues, the quality of shifts can become a visible aspect of employee experience as much as pay.

Why should managerial control be maintained?

One of the healthiest frameworks in Deputy’s guide is positioning AI as a helper rather than an "autopilot". The system can make predictions and draft plans; it is suggested that the final decision remains with the manager.

There are several reasons for this.

Data looks back, while the manager may know specific situations that will occur in the future.

The past three Tuesdays may have been calm. However, if there is a 400-person conference in the nearby area next Tuesday, a model based solely on historical data may predict low demand.

Similarly, someone on the employee list may appear "available", but the manager may know that the person is not yet capable of working independently at a specific station.

Thus, operational knowledge is still needed to approve the AI's suggestions.

The legal and ethical aspects are growing

When AI is involved in workforce management, the issue goes beyond efficiency.

Current AI regulations in Europe evaluate some AI systems used in employee management as particularly sensitive areas. According to a legal analysis by Reuters dated July 2026, areas such as job allocation, monitoring employee performance, and AI applications affecting promotions or the continuation of employment relationships may be considered high-risk usage areas. The human manager's final decision does not automatically eliminate risk in situations where it significantly affects the decision-making process of the system.

Therefore, businesses should not only ask "Is there AI?" when purchasing technology.

More accurate questions include:

What data does the system use?

Does it score the employee?

What criteria does it use to suggest shifts?

Does it affect performance decisions?

Where is the employee's personal data stored?

Can the manager see why the suggestion was changed?

How can incorrect decisions be corrected?

Especially in multinational hotel and restaurant groups, these questions become issues that HR, legal, IT, and operations need to evaluate together.

Where should a small business start?

The first step to transitioning to AI-supported workforce management may not be purchasing expensive software.

The current shift plan of the business should first be measured.

Sales can be compared with staff hours over a four or eight-week period. Days and times when there were staffing shortages or excesses can be examined. Overtime, last-minute leave requests, and shift changes can be recorded.

Then a small pilot can be conducted.

For example, the staffing needs suggested by the system for just Friday and Saturday nights can be compared with the manager's estimates.

Three questions are sufficient:

How close was the AI prediction to reality?

Which suggestions did the manager change?

What was the reason for the changes?

These records will reveal over time whether the system truly creates operational value.

The AI plan should be shareable with employees

Another risk is explaining shift decisions with the phrase "the system said so".

If an employee's hours are reduced or their shift is changed, management cannot shift responsibility to the algorithm.

Even if AI generates suggestions, the decision belongs to the business.

Therefore, it may be valuable for employees to be informed at a basic level about the system used, to explain which data affects planning, and to have an appeal mechanism for decisions that seem erroneous or unfair in terms of employee trust.

SHRM also emphasizes that the use of AI in the workplace should be supported by employee training, transparency, and human-centered practices.

Stafflink’s view

The biggest opportunity for AI-supported shift management in HORECA businesses is not to work with fewer people, but to reduce the problem of having the wrong number of employees at the wrong time.

In our view, setting up the system solely with the goal of reducing labor costs may look good in the short term but can lead to results that may harm service quality and employee experience.

A stronger model should optimize three data points simultaneously:

demand, employee competency, and employee sustainability.

The cheapest shift on a Friday night may not be the most efficient one.

The true value of technology will emerge here. If it provides better predictions for managers to make decisions, reduces last-minute uncertainties for employees, and maintains service quality, AI is a meaningful operational tool.