Potential of Predictive and Video-Based AI Solutions in the Food Industry

Jörg Brezl: Any company, regardless of size, can benefit from our products. That's because the solutions offer a wide range of possible applications in all aspects of the production chain. As a software company, we support both small and medium-sized as well as large companies in the meat industry. That's why we know exactly what challenges the industry is facing. Some of the largest are:

• • Shortage of skilled labour
• • Sustainability
• • Declines in sales and revenue
• • Lack of transparency
• • Pricing pressure
• • High use of resources
• • Changing market conditions due to dynamic consumer trends
• • Cost-effective storage
• • Insufficient progress in digitalisation.

Jörg Brezl: Our AI solutions can usually be assigned to one of the following three categories, which directly or indirectly influence the problems mentioned:

• Ensuring animal welfare and animal health during transport and slaughter, which has a direct influence on food safety and meat quality.
• Automation in cutting, packaging and shipping, enabling reliable quality control.
• Predictive AI solutions for efficient planning in production, purchasing and logistics, thereby reducing food waste, resource consumption and lost sales.

Jörg Brezl: I'll be happy to outline a typical use case. The first cattle transports arrive on Monday. The situation used to be as follows: All the animals move around while the trained staff try to count them. Has this animal already been counted or has it just moved to a different position when it was briefly noted on a sheet of paper? But now the company has an AI solution from SLA that automatically counts the animals using the "Classifai Box2 and detects possible injuries. This relieves the burden on employees and helps them to reliably record each animal. At the same time, the data is directly available in digital form.

The next step is the slaughterhouse. Whereas in the past, for example, it was laborious for specialists to read and enter or confirm the sender ID, slaughter number or sex, AI now reliably supports them in these tasks. This enables trained staff to devote themselves to other tasks such as quality or slaughter hygiene. AI also records animal welfare characteristics, which serve to improve transparency and husbandry conditions. In addition, the assistance systems can help with organ diagnostics, which supports and relieves the burden on specialist staff during meat inspection.

And a lot has changed in this company when it comes to cutting and packaging as well. That's because AI-supported product recognition, combined with simultaneous, automated weighing and labelling, is relieving the burden on skilled workers, who are increasingly able to devote themselves to quality control. In addition, the image-based AI solution reliably and accurately identifies foreign objects. In the next step, AI also provides support during delivery. She records the product numbers of the products, for example sides of pork, that are loaded into the truck. This means that the company knows exactly which products its customers receive, enabling it to ensure product quality and traceability.

Jörg Brezl: This solution simultaneously supports the company's production planner. This is because this person has to decide on the same Monday morning what and how much is to be produced, even though the orders don't come in until the afternoon. So what basis should he use to make a decision? He used to rely on his many years of experience and consider which factors could influence the product quantity. He also researched whether a customer is planning a discount campaign, what happened in the past at this time of year, etc., and what was decided. However, this is associated with a great deal of effort and, in the worst case, leads to too much being produced and these products then being sold at too low a price. To avoid food waste, SLA's predictive AI solution is the optimal support at this point.

Based on a wide range of data and influencing factors, the AI module provides assistance in making predictions that can be used as a basis for forward-looking and efficient decisions. Furthermore, SLA's predictive AI and business intelligence solutions can even be used to determine how accurate the predictions were, how large the margin was and whether the AI module needs to be optimised. The AI module that makes predictions is also a great help with another aspect: What happens if the production planner goes on holiday or is absent due to illness, or retires? Often, the necessary expertise is stored in the heads of such production planners. But SLA's fully integrated, predictive AI solution makes this knowledge more transparent, so that the substitute or successor can continue working immediately.

Jörg Brezl: While the words "reliable" and "fast" are usually considered to be in conflict, our AI solutions manage to combine these contradictions. They are fast, precise and reliable and can be fully integrated into existing operational processes with a minimal investment. Personnel costs and training expenses for certain processes can be saved.

Whether it's the more reliable assignment of animals, findings and origins, the recognition of animal welfare characteristics, the collection of data at slaughter, the improvement of product quality, predictions about resource consumption (e.g. water), better transparency and traceability or in production planning – SLA's AI solutions minimise errors, ensure the completeness of all data, relieve the burden on specialists, help to conserve resources and avoid overproduction and with it food waste and lost sales.

This means that they offer the meat industry considerable potential for savings and optimisation, which contributes to both the efficiency of the supply chain and to increasing profitability. By automating processes and integrating predictive analytics, we enable more efficient, precise and sustainable production and shipping logistics. It should also be emphasised that SLA's AI solutions can be integrated into any ERP system and are therefore useful for any company.

Jörg Brezl: AI will change the labour landscape in the food industry in the medium and long term. But that doesn't mean that it will replace skilled workers, because they will now take on new areas of responsibility. Our aim is to relieve the burden on existing trained staff and to support them. By automating tasks, our AI components enable employees to focus on other tasks. For example, AI is used to automatically count and monitor animals, a task that used to be done manually. The AI components also act as an assistance system when inspecting carcasses. While these tasks used to require precise work under extreme time pressure – which often led to stress and errors – the expert staff can now concentrate more on quality assurance, hygiene controls and the processing of special cases. This not only improves product quality and efficiency, but also working conditions, which in turn increases employee satisfaction and productivity.

Nor do we leave our customers and their skilled personnel alone when it comes to using the AI modules. Instead, we train and educate the company's skilled staff and involve them in every step of the automation process. That's because we've learned one thing: regardless of which ground-breaking technology is introduced in the company, it will only work if the employees are on board from the outset.

Jörg Brezl: Due to the low upfront cost, our AI solutions often pay for themselves in a matter of days or a few weeks, depending on the specific application and operational circumstances. In slaughtering and deboning, AI-supported systems lead to a very rapid improvement in quality assurance and product quality. As these processes are directly integrated into the production process and have an immediate positive impact on efficiency and profitability, the benefits of AI solutions are often noticeable within a short period of time. One example of this is the reliable use of our image-recognising AI solution for the digital sorting of outgoing items in the cutting plant. In combination with checkweighers, the AI solution recognises the correct article in the disassembly output and, by linking it to metadata, even recognises the quality. By reducing manual intervention in this repetitive and time-consuming task, sorting errors can be avoided and the deployment of personnel can be significantly reduced.

Investments in predictive AI solutions often pay for themselves within a short period of time. A typical example: Let's assume that each kilogram of meat is sold for € 5 per kg, and a company produces 500 tonnes per day. Without accurate forecasts, about 5 tonnes could be produced unnecessarily, resulting in a potential loss of revenue of € 25,000. If our predictive AI solution improves the accuracy of the forecasts by 10 %, this loss can be significantly reduced.

Jörg Brezl: Our AI solutions require a very small investment and minimal hardware compared to conventional solutions and can be easily implemented in all existing systems. Before our customers choose a suitable AI solution, however, a few important preliminary considerations and preparations are necessary to ensure that the implementation is successful and delivers the desired added value. The most important questions include:

• To what extent has the company's data already been digitalised and is it already complete?
• What technical infrastructure, in particular network infrastructure (LAN cable, internet connection, etc.), is available on site?

We support our customers in these preparations by conducting joint analyses and on-site visits. This enables us to work with the customer to evaluate their specific situation and determine what preparations are necessary and which AI solution best suits their needs. We then support the entire process, from planning and implementation to optimisation, to ensure that the AI solution delivers maximum benefits.

Jörg Brezl: We rely on both newly generated data from us and existing digital data from the customer to ensure that our solutions work accurately and efficiently and achieve the best possible benefits for the business. There are two options for AI solutions based on image and video analysis: The first is that we create the required training data by picking up images and videos directly at the customer's company. The second option is our "SLA AI Labelling Platform". Here, the customer simply registers and can upload, label and annotate their data, as well as directly view, test and validate the results of its selected AI components.

We also use various types of historical and current operating data from the customer for our predictive AI solution. Historical ordering data is particularly valuable because it provides insights into past sales and production patterns. As many of our customers already have ERP systems in which this kind of data is systematically collected and stored, we can easily use this data to train the predictive AI solutions. The additional data we use for training also includes several influencing factors such as holidays or weather, as they have a direct influence on demand.

Ideally, we also receive information about planned price promotions in the food retail sector to which our customer delivers his products. But specific information on the size of the food markets is also important, as these determine the purchasing figures. However, this data is often not well structured and digitised. If this is the case, we support our customers in digitalising their data to create an optimal basis for training. The basic rule is: The more of the customer's data that is available in digital form, the faster initial results can be achieved.

Jörg Brezl: It will be revolutionised to a significant degree. We are already seeing far-reaching effects that the use of AI has on the entire industry. For example, the trend is moving towards multi-component solutions that make it possible to seamlessly connect various processes along the production chain. These range from slaughtering to deboning and packaging, and thereby further increase efficiency and profitability. The use of reliable AI and automation solutions also enables foreign bodies to be detected with precision, errors to be reduced, processes to be optimised and product quality to be improved across the board. We are already seeing changes in the way resources are used and in animal welfare. This is where our AI solutions are increasingly helping to make the meat industry more sustainable: Precise forecasting allows production requirements to be planned more accurately, preventing overproduction and food waste. At the same time, continuous detection and analysis of animal welfare characteristics enables deviations to be detected at an early stage. That in turn leads to improved husbandry conditions and better meat quality. The use of AI also reduces the workload of skilled workers, who can now focus more on quality assurance, customer service and other important tasks. This is a significant economic factor, especially in times of a shortage of skilled workers.

In catering and commercial kitchens, our predictive AI systems can be used, for example, to optimise purchasing and production processes. They help to predict demand more accurately, which in turn reduces food waste and optimises inventory levels. In the hotel industry, they can help optimise operations by, for example, predicting the consumption of food and other resources and managing purchasing accordingly. Other areas in which we're active include agriculture, veterinary medicine and animal welfare, as well as logistics and transport, food retailing and production in the food industry as a whole. With various projects and our advanced AI solutions, we are continuously working to expand the applications of our technologies to new industries and to develop customised solutions for different market requirements.

Classifai Box and Classifai Box Plus are the central hardware components of SLA's image- and video-based AI modules, which are already being used by numerous customers. Thanks to the extremely low hardware and investment requirements combined with a wide range of possible applications, this AI solution is very successful.

One of the most important applications of the Classifai Box is the automatic inspection of carcasses, based on which recommendations are made. The AI modules recognise and classify relevant data such as different body parts, markings, injuries and illnesses in real time. A major focus is currently on the recognition of sender codes, slaughter numbers, ear tags, gender and boar risk, as well as on the classification of pig tails and meat and fat portions in cattle. In addition, animal health and welfare characteristics such as tail length, partial losses and pathological changes to ears and organs are reliably identified. This not only contributes to quality assurance, but also to the promotion of animal welfare. In addition, precise allocation of the animals and their findings ensures traceability. In the cutting plant, the Classifai Box automatically identifies at least 500 different meat products in boxes and is connected to scales and labellers to automate several work steps.

SLA Software Logistik Artland GmbH
Jörg Brezl, Geschäftsführer
info@sla.de
T +49 (0) 5431 9480-0
https://sla.de