How MAIF cut modeling time and built thousands of models with Akur8

+90k models and model versions created

+500 databases uploaded with up to 30 million rows

+5.5k grid searches generated

––

Challenge

MAIF is a leading French mutual insurer, known for its strong presence in motor, home, and personal lines insurance. With pricing models reviewed annually, the team operates under tight timelines and a high volume of work. These constraints led them to explore new ways of working and consider a modern actuarial platform.

Their primary objective was clear: reduce modeling time without compromising performance or model quality. The second priority was improving collaboration. Given that many modeling tools are designed for individual use, MAIF required a team-oriented solution that would make it easy to collaborate and share models seamlessly.

At MAIF, we review our models every year. Just for home insurance, we manage five different contracts, and for each contract there are about 10 coverages to model within a fairly short timeframe. So the first big challenge was to be able to save time on the modeling process, but without losing performance and model quality.
Anne-Sophie Dyèvre
Actuarial Research Lead

Main challenges

  • Tight timeliness and a high volume of work
  • Fragmented systems that slow down the processes
  • Poor collaboration across pricing team

––

Solution

For MAIF, the first major shift with the Akur8 rollout was bringing every step of the modeling workflow into a single tool, from data preparation to model building and geographical modeling. This replaced a fragmented setup and made the overall process more consistent and easier to manage.

Previously, MAIF had to switch between modeling and programming tools such as SAS or Python to prepare, adjust, or refine datasets. With Akur8, these changes can be made directly within the platform, making common tasks much quicker.

On the modeling side, Akur8 streamlined the workflow by automating the repetitive parts, while keeping actuarial expertise at the core. Our proprietary Transparent AI helped the MAIF team save substantial time by exploring thousands of variable combinations in parallel and identifying the most predictive features, while preserving transparency and keeping actuaries fully in control of final selections and judgment calls. This balance strengthened model performance without compromising interpretability, governance, or expert decision-making.

Akur8 simplifies modeling while still leaving us in control of our choices and interpretations. It also gives us flexibility to adjust the models, even though a large part of the process is automated.
Anne-Sophie Dyèvre
Actuarial Research Lead

Implementing Akur8 also transformed collaboration. By consolidating work in one integrated platform with shared projects, MAIF’s pricing teams can work together more easily. Multiple users can contribute to the same model and share projects across the team.

Key results

  • Thousands of models and model versions created
  • Centralized processes that helps to move faster and are easier to manage
  • Improved collaboration across pricing teams

––

Conclusion

This centralized setup has helped MAIF move faster. They reduced modeling time without compromising performance or model quality, while improving collaboration across teams.

Today, across all lines of business and pricing initiatives, MAIF manages several hundred databases ranging from a few million to 30 million rows, as well as several thousand models and versions, all within a single platform.

Their transformation doesn't end here. In the coming months, MAIF plans to test the Rate module to explore the move from pure premium models to technical rates, and to benchmark those rates against those currently in production.

Ponte en contacto con nuestro equipo de ventas para obtener más información sobre la solución.