Digitalisation is omnipresent. From logistics to healthcare, AI and machine learning are reshaping industries at an unprecedented pace. But what does this mean for cheesemaking – an industry steeped in tradition, craftsmanship, and sensory expertise?
At first glance, cheese production may seem immune to digital disruption. Yet, beneath the surface, it’s a process rich in variables: milk composition, temperature, pH levels, moisture content, and microbial activity – all of which influence the final product. This complexity makes it a prime candidate for digital innovation.
Here are Tetra Pak’s Tim Dijkstra, Solution Manager Analyze, and Joanna Ilczyszyn, Commercial Manager, to give you a rundown of the possibilities new tools and technologies open up for cheesemakers – and to give you a sneak peek of the future.
Step one starts here. With a challenge.
At Tetra Pak, we have deep domain expertise that extends beyond equipment and services. Our team, which we call Analyse, harnesses this knowledge to transform customer plant data into valuable insights that drive efficiency and reduce costs. By applying our industry-specific understanding, we add essential context to raw data, creating powerful applications tailored to the unique needs of food and beverage manufacturers. To achieve this, the first step is to ensure a clear problem statement and the availability of data.
“We ask our customer ‘what is your challenge?’. And then, ‘do you have data related to that problem?’. And if you have positive answers to those two questions, there's a potential to look into the possibilities. We say possibilities – because there will be some challenges. We will, for example, face a diversity of data sources. Some data will be on paper and some will be stored in different databases across the premises. To unlock the value of data, you want to bring it all together in a single source of high-quality data. We can support customers by upgrading their infrastructure – we’re building a future-ready infrastructure as a standard platform already today,” Tim says.
Step two. Solution time.
With the prerequisites – infrastructure for data collection – in place, how can that data be used? Tim again:
“Our scope encompasses diverse competencies and capabilities – cheese technologists, data scientists, automation engineers, and, of course, there are also our customer-facing people to transfer the message. So it's quite a multi-discipline area.”
To explain what a ‘scope’ can comprise, Tim exemplifies with an AI model for optimisation of one of the main quality indicators in cheese: the moisture level.
“We have developed an algorithm that lets customers control cheese moisture levels. To implement this, we first analyse whether there is a valid business case for the specific customer. For that, we need access to a specific set of data. Interpreting this data requires both a data scientist to analyse it, and a cheese technologist to explain its meaning. So the interpretation of data is a joint effort. With our algorithm and the customer’s data, we build the model, validate it, and deploy it. Once deployed we continue to improve the model by updating the data set, analysing the performance and adapting the model accordingly. There is also an interface for the operator, and we train them in how to work with it. It is a complete package.”
This enables customers to optimise and achieve complete control of the moisture levels in their cheese. It builds knowledge around the optimal levels and values. And the same kind of workflow can be used to control recipes and other parameters. Joanna Ilczyszyn adds:
“The process we are using for the moisture optimisation tool is a good example of how it's happening on site. It's about digitalising manufacturing processes, and machine learning is a super important part of that,” she says.
Step three. Looking ahead – to autonomous plants
Cheesemaking is a craft. You need experience to become really good at it. This means that cheese producers are dependent on the knowledge of individuals. In times of skills shortages, this is a challenge – which can be addressed through automation and digital tools. And while AI and machine learning are not objectives in themselves, these phenomena are inevitably playing an increasingly significant role.
“Today AI is an assistant, something that helps the cheese experts in their daily work,” Tim says.
But what about tomorrow? The journey onwards is continuous and has no fixed end destination. But as far as the near-future roadmaps go, the direction we are heading, step by step, is towards the concept of autonomous plants. Joanna explains:
“We aim to deliver a full understanding of the cheese line, which is complex in terms of number of units, operations required, control factors and potential risks. So how will we be able to achieve this? By foreseeing what you need to do from an operational perspective, what you need to improve and what you need to avoid.”
But what about tomorrow?
The journey of digitalisation in cheese production is far from over. In fact, it’s just beginning. The road ahead is continuous, with no fixed destination – only evolving milestones. And as we look toward the near future, one concept is becoming increasingly clear: autonomous plants.
“The direction we are heading, step by step, is towards the concept of autonomous plants,” Joanna explains. “Autonomous plants represent the next frontier, where systems not only monitor and report but also learn, adapt, and act independently. Imagine a cheese production line that adjusts itself in real time based on incoming milk quality, environmental conditions, or even market demand. That’s not science fiction – it’s the future we’re building.”
The foresights are based on trends in the collected data, which in turn is based on decades of accumulated know-how.
“The design of AI is based on the logic of a human brain. So if an experienced cheese technologist can make perfect cheese, AI can mimic that, at least to a certain extent,” Tim says.
Joanna continues:
“So we take the knowledge the human brain has fed to the production system, and we use that knowledge to stabilise the parameters that have an impact on revenue. We can look at media consumption, or the CIP processes, or the efficiency of the vat … by making sure our equipment has the proper software and hardware to cover all data points that are influencing line output, we’re automating production and ensuring that product quality and revenue do not depend on decisions of the individual operator.”
For new factories, where cheesemaking experience may be limited, the potential is huge. Joanna elaborates:
“You can't teach someone to become a cheese technology specialist in just a year or so. It's impossible. But what is possible, is to utilise knowledge from data collection, data understanding, and data structure, to create a logic for the process flow and then apply that to the AI model, which is then reading, controlling, and analysing data on the fly in the line – like an operator.”
Experts will be experts
But when technology steps in, what happens to craftsmanship and expertise gained through long experience? Is it all lost? Tim flips the perspective over:
“On the other hand, what happens when an expert leaves or retires and the successor does not have the same knowledge? Now, you can use the AI model to train the new technologist and retain expertise.”
It is, however, very much a step-by-step transition.
“When you implement AI, generally only 25% of the budget goes to the creation and implementation of the AI model. The rest, 75%, is change management – how do you get people to accept this new technology? That’s where the long-term relationship we have with our customers come in. We don’t just implement new technologies and say, ‘good luck with it’. We make sure it will work, and we’re with our customer on their road forward,” Tim says.
Human knowledge is still very much the foundation the processes are built upon. When it comes to producing that perfect cheese, you just cannot beat an experienced cheesemaker.
“AI cannot do the job better than a human cheese expert at his or her best. But a human gets tired, a human needs to eat and drink and so on. AI is consistent. It makes your cheese in the same way every time. But again, it is important to underline that people are always in charge and people make the decisions,” Tim adds.
Joanna exemplifies with the coagulation of cheese.
“In the past, the operator would open the vat, squeeze the coagulum in their hands and look at the structure to decide if the coagulum is ready to be cut. Today, coagulation can be controlled via sensors in the equipment, measuring the moisture level. So over time, AI helps us analyse data in a way that lets us evaluate parameters and control performance better. But we will still need the human eye, experts saying ‘yes, this trend is correct, this will work’.”
The future is, per definition, hard to predict. But regardless of what it may have in store for us, Tim’s and Joanna’s teams, with their comprehensive, one-stop-shop expertise, will continue to support the cheesemakers of today and tomorrow; continue to fine-tune, improve and optimise. Because, as Tim puts it:
“There’s never a situation where there’s no need to improve further.”
