With the rise of ChatGPT, Artificial Intelligence (AI) and Machine Learning (ML) have made their way from the tech world into the mainstream. But even though the broader public seems to have just discovered AI, many companies have been working towards AI and Machine Learning solutions for quite some time as it proves to be a crucial step in building a future proof business. 

At HelloFresh, we’ve been heavily investing in AI and ML for more than six years now and see them as key components of our technology platform. We currently have many AI driven use cases that provide real monetary benefit and help us to become an even more efficient business. With more than 70 data scientists and machine learning engineers worldwide, we have a team of highly skilled experts that dedicate their work to continuously training, refining and deploying machine learning tech tools. This article will give an overview on how AI and ML solutions are already making HelloFresh a more efficient business and offer an outlook on the potential generative and predictive AI entail. 

Data from 12 years of HelloFresh feed into more than 1500 models being trained and deployed per week

Data plays a very important role in the entire meal kit production process. With more than 10,000 recipes, 12 years of customer order patterns, and menu browsing behaviors, HelloFresh has a very valuable and unique database. Since its launch in 2011, the company has grown the largest and richest customer database of taste preferences worldwide. A stable data infrastructure is the base for making use of this unique data. With seven brands across 18 markets, there is a huge amount of data being generated on a daily basis. AI and machine learning models rely on consumable data sets and a self-serving data platform which our tech team continues to evolve in order to make use of the valuable data we have. 

“In the past few years, we’ve been heavily investing in building up a team of talented scientists and engineers to set up the base for the HelloFresh Machine Learning models. I’m incredibly proud of the great projects that have been launched since I joined HelloFresh almost two years ago”, said Val Liborski, CTO at HelloFresh.

There are a number of AI-driven applications that HelloFresh is already actively using across marketing, operations and product teams. Most of these scaled use cases rest on predictive AI, while we are looking forward to unlocking the potential of generative AI models in future. 

AI driven Menu: Machine Learning based product recommendations

AI driven menu recommendations will further individualize our customers’ product experience. 

In the long-term, Artificial Intelligence will enable a fully personalized product experience for HelloFresh customers worldwide. In the US, customers can already experience that: preselected options are now ranked in the menu by a machine learning algorithm trained on customer's previous meal selections in order to display the most relevant products for them, first. This is the first machine learning based solution we’ve launched to individualize our customer’s menu selection. 

Furthermore, AI and machine learning offer a big potential for optimizing our HelloFresh Marketing costs. One of the areas we are already actively using AI are customer value predictions. It enables predicting future actions of our prospective and active customers, we can individualize marketing for different customer groups and optimize and personalized discount offers. By recommending and distributing advertising budgets across creatives, audiences, channels and markets, we can maximize the return of such marketing interventions and accelerate growth.

The potential of AI for the future at HelloFresh

Beyond the exciting applications we already have in place, HelloFresh is experimenting with many more AI and machine learning models that will make the business even more efficient and sustainable in the future. In fulfillment, we are currently testing a machine learning driven packaging optimization algorithm. With the help of weather forecasting and packaging material data, the algorithm recommends dynamic packaging solutions that will lead to savings on packaging material – a big opportunity for decreasing our environmental impact and costs across the value chain.

Furthermore, we see a big potential for AI to improve productivity across all teams at HelloFresh. Within the tech teams, AI can make coding more productive. In the long-term, generative AI models can support our video and photo production, significantly saving costs and supporting our business at scale. 
“Given our rich experience and the talent within our tech team, we are well equipped in unlocking the full potential of AI in the years to come,” Val comments.