Search & Findability Business Consultant, Manager at Tietoevry
The challenge
Systembolaget wanted to explore how Artificial Intelligence (AI) can be used to define and develop a data model for similar wines and future digital services.
The hypothesis: An AI-solution can recreate and develop domain-specific expertise on similar goods, thereby providing relevant suggestions for "similar wine"
The solution
Starting the project we did not know the quality of the data, the possibility to create an AI-model that can read that data, what ability Systembolaget had to train such a model, or how such functionality can be presented in a front-end.
We worked together with Systembolaget, in an agile and cross functional team using an iterative approach. In a limited timeframe, we delivered an AI-driven application which recommends similar wines based on taste.
About the customer
Systembolaget’s mission is to sell alcohol beverages responsibly while offering first class service. It is a retail monopoly not driven by profit or sales but rather based on Sweden’s restrictive alcohol policy which aims to limit the harmful effects of alcohol.
Read more about SystembolagetMain benefits
Knowledge of AI-technology
Today Systembolaget has a far greater knowledge in “How to use AI and machine learning to solve business problems”. This knowledge can be found on all levels in the organization
Better recommendations to customers
The final AI-application is used in stores supporting personnel in their advisory role and giving recommendations on similar wines as part of the digital customer experience in the mobile app.
Helping customers find similar wines
Wine suggestions are based on similar taste and aroma patterns. Combined with filters on region, price, grape or food; possibilities for further development of recommendations to customers are endless
Systembolaget wanted to learn how they could use AI to solve real business problems increasing their service level towards customers. Together with Findwise, part of Tietoevry Create, Systembolaget has taken the idea of AI driven recommendations based on taste all the way into production.
The function uses an Artificial Intelligence model to give the customer suggestions for new wines with a similar taste as their favorite wine. When the customer has specified a wine whose taste they like and asks for suggestions for similar wines, suggestions are given on other wines with similar taste and aroma patterns. The products proposed are of the same size so that the customer is not tempted to buy more than intended. The final AI-application is used in stores supporting personnel in their advisory role and giving recommendations on similar wines. It has also been piloted as part of the digital customer experience in the mobile app.
We know that many customers find it difficult to choose wine. Now we are trying to reach out with our advice in a new way. It will be exciting to see how it is received by our customers
Frida Jarlbäck
Business developer at the Strategy and Offering department at Systembolaget
The function Similar wine, has been developed with the help of Systembolaget's beverage experts, evaluated by their knowledgeable store staff and available for customers to try this new way of getting beverage advice.
From a Proof of Concept to production
The project started out as a proof of concept to explore how AI can be used to define and develop a data model for similar wines and future digital services. Findwise worked from the beginning together with Systembolaget in an agile and cross functional team. Using an iterative approach we delivered an AI-powered application which recommends similar wines based on taste.
The result of the PoC was very successful and Systembolaget chose to go further with a production version in their labs concept, first as a tool for store employees and now as a service for the end customers.
Please contact us if you want to know more about the technology used, learn to rank in Elasticsearch using a decision tree model, or if you want to learn more about how we brought a successful AI-PoC into production using an AI-as-a-service model in a public cloud environment.
