Inside Tetra Pak’s AI Journey: Interview with Alberto Barroso

As AI continues to reshape industries and organizations, the world seeks insights from pioneering companies that have boldly embraced AI to learn how to navigate its complexities. Recently, we had an incredible opportunity to sit down with four industry leaders at AMLD EPFL 2024 and talk about their perspectives, thoughts, and experiences on structuring and designing winning AI programs. 

This time, we're bringing you a piece of our conversation with Alberto Barroso from Tetra Pak. Alberto shares his experience, detailing how Tetra Pak began their AI journey, the challenges they encountered, and the key lessons learned along the way.

Can you tell us a little bit about Tetra Pak?

Alberto Barroso: Tetra Pak is a company focused on packaging food manufacturing, with around 25,000 employees dedicated to making food safe and available everywhere. Our mission is to protect what's good and lead the industry in sustainability efforts.

What is your role at Tetra Pak?

Alberto Barroso: I’m Head of the Decision Science Centre, a unit focused on optimizing decision-making by integrating both data-driven evidence and human insights.

How does Tetra Pak leverage AI in its operations?

Alberto Barroso: We leverage AI internally to optimize our supply chain, sales process, and pricing strategy. We also use it to develop new products and services, such as making our machines more efficient and sustainable with less energy and water consumption. This supports our commitment to achieve net-zero neutrality by 2030 and zero value chain emissions by 2050. 

Sustainability is at the heart of our operations. We invest heavily in renewable and carbon-neutral packaging. For instance, last year, we eliminated the aluminum barrier from our packages to reduce waste while maintaining food safety. As the industry shifts towards paper-based packaging and faces increased competition from online retailers like Amazon, we focus on optimizing packaging content and advancing sustainable practices.

How did Tetra Pak’s AI Journey begin?

Alberto Barroso: In 2018 we established a Decision Science Center of Excellence. Our goal was to focus on practical use cases to optimize decision-making across various domains. By 2019, we found our first scalable use case: optimizing cheese production. Using machine learning methods tested in randomized trials, we saw significant improvements in cheese moisture and quality. This was a crucial step forward. Then in 2020, we intensified our efforts. We adopted the THEARI framework, which was released during COVID, addressing the challenges scientists faced in sharing and understanding evidence. This framework helped us reflect on our methods and emphasized the importance of standards for backtesting, randomized clinical trials, and continuous performance control—essentially applying a scientific process to AI development.

Over the past year, we’ve been really focused on this program called Accelerating AI. This program aims to use specific cases to influence decision-makers' behaviors, and it has helped us to add more use cases and see how they impact our organization. We’ve noticed that people are struggling with acting based on evidence, understanding scientific evidence and statistics, and we feel like there is a need to create frameworks that incentivize rational decision-making.

Thanks to this program, we now have 19 use cases, with 13 already proving value using scientific evidence, 11 undergoing clinical testing, and three in production. Our goal now is to bring transparency to the organization about how this works. Understanding where we have proven evidence is crucial to bringing change and a different way of decision-making.

Every area of the organization, from finance to HR to supply chain, procurement, services maintenance, and processing solutions, is involved in these use cases, and we’re trying to build trust by discussing them and presenting evidence to everyone in the organization.

AI has significant potential, especially with Large Language Models (LLMs), but we must ensure we use these models correctly. We have to ask ourselves: “Is this a good tool, what value does it provide?”

What key learnings have you gathered from your AI journey? 

Alberto Barroso: Here, I’d like to highlight our Accelerating AI program, which annually brings around $900 million of opportunity value, through cost reduction and optimized decision-making. This initiative has shown us that AI can help develop new business models. And by just collaborating with people across different units, we’ve been able to expand 19 projects into multiple use cases, such as optimizing pricing, total cost of ownership, supply chain, and procurement.

There is a significant cross-fertilization within these programs, but the major challenge we face is prioritization. As a company, we try to make decisions efficiently, yet changing the way we work is difficult. For example, each time we prove the value of AI, we aim for faster testing and larger randomized control trials. However, we often lack this process in our DNA. So we are putting in a lot of work into integrating KPIs to prioritize decision-making differently. 

Lastly, customer focus is essential. Creating value with AI involves not only understanding data, but also understanding how people react to and use this data to make decisions. 

For more insights into the AI journeys of leading companies, such as dsm-firmenich and TE Connectivity, check out the entire panel discussion here.