CEOS, the first artificial intelligence platform serving business specialities
The 27 candidates from a number of departments were able to get their hands on CEOS for the first time during a Data Challenge. The goal? Designing a solution based on artificial intelligence for optimising LifeTime to Value* for new clients.
CEOS is an environment that lets us deploy machine learning models. Depending on the parameters learnt, these artificial intelligence systems are, for example, capable of optimising a client's value: after analysing the characteristics of our client files, AI is able to target the best offer (NBO) that our representatives can put forward.
This platform designed for data science represents a technological innovation at CA CF. Hosted by the AWS cloud on an open-source Kubernetes foundation (a solution that ensures efficient and secure orchestration of the applications deployed), CEOS adapts to fluctuating demand and to the market: the platform autonomously arbitrates decisions on adding and decommissioning resources depending on requirements. It manages itself.
This first full scale test has been a very conclusive one, opening up new horizons for more advanced AI usages. Further, after this challenge, three new departments, Credit France for granting loans and fighting fraud, the Data Management Office and Risks & Continuous Monitoring and Group Compliance unit could all soon benefit from this platform.
From an organisational point of view, at CA CF, the Digital Services France teams are the prime movers. They are using a so-called MLOps** approach,
- The Architecture and Innovation team provides the business tools and the industrialisation processes for the AI models.
- The Data Tools teams sets out the development framework and supports the specialities in onboarding the new environment.
- Lastly, CA GIP makes the technical base available and maintains it.
To find out more about CEOS, don't hesitate to contact Marine Leplat mleplat@ca-cf.fr, Martial Leroy mleroy@ca-cf.fr or Samuel Dehouck sdehouck@ca-cf.fr.
* LifeTime Value is an indicator used to identify the total profits generated by a client over a given period of time. This way, Sofinco could, right from the start of the relationship estimate the profits brought by the client and propose the most suitable offers both for them and for Sofinco. Hence, Sofinco could optimise its sales investment by concentrating on those clients that generate more revenue as well as relaxing the acceptance criteria for approving loans to these same clients.
* * MLOps (Machine Learning Operations) is an approach that aims to integrate and automate the deployment, management and oversight process for automatic learning models. It facilitates collaboration between development teams and Data Scientists while allowing model maintenance in an effective and evolving manner.