Scania built a parts-pricing AI in 48 hours and put it into production four months later

Scania built the foundation for Nexplex, an agentic AI system for improving spare-parts pricing, during a 48-hour internal hackathon. Four months later, the system was in production, according to Swedish publication Techtidningen.

Scania handles millions of spare-parts transactions each year, and prices depend on the cluster assigned to each part. To check whether a part is classified correctly, staff must combine information from technical descriptions, purchasing systems and workshop history. Techtidningen reports that Nexplex uses data and AI alongside business rules drawn from experienced employees’ knowledge.

The project team had to work with senior experts to turn that experience into rules the agent could use. Tushar Kale, a Solutions Architect in Scania’s Data & AI Foundation, said this took workshops and substantial effort; the rules were not already collected in a single document.

The hackathon team included 22 people and was the largest among around 500 participants. During the two-day event, it used Snowflake Ontology to build on data already gathered in Snowflake from ten sources. Kale said the application covers about ten datasets, 150 million nodes and 480 million relationships.

Techtidningen says Nexplex has already contributed significantly to increased profitability, but provides no figures. The article also describes continuing challenges in obtaining additional data as the system’s needs grow.