The challenge now for companies is no longer (only) that of the design of artificial intelligence models and their experimentation. It is that of their industrialization, thanks in particular to MLOps approaches.
The publisher Cegid also wishes to increase its use of AI, embedded in its software solutions or for its own business needs. To this end, the French group is striving to multiply the development of AI models, while shortening design cycles.
No code for creating AI models
This is precisely the promise of DataRobot, of which Cegid is a customer of the platform. This aims to automate “the complete analytics cycle, including the development, deployment and ongoing management of models”.
DataRobots is notably used to design predictive models intended to predict the probability of paying bills and the propensity of customers to add services.
“We wanted to reduce the need for coding while accelerating model development,” says Joao Henriques, head of credit risk management and financial analysis.
The team in charge of invoicing therefore resorted to DataRobot’s no-code functionalities. For its integration and configuration for decision-making, Cegid called on Passio Consulting.
The AI platform is integrated with Cegid’s AWS IT environment and its data lake through APIs. Data is rendered within Microsoft PowerBI and Excel.
+ 20% of factoring contracts without increased risk
The first use case concerns factoring. Cegid offers its customers a financing solution. To make these credit decisions, the publisher has designed a “one-click process” based on the analysis of customer data and risk scoring.
Using a payment forecasting model (PPM), the company assesses the probability of payment for each invoice. Data and prediction also help him set optimal interest rates.
“We use machine learning models to support decision-making on all the individual transactions we carry out in our factoring business,” explains Joao Henriques.
Thanks to machine learning, Cegid estimates a 20% increase in the approval rate for factoring contracts at the same level of risk. This increase generated 15 million euros in additional annual revenue.
The credit risk department is now considering adopting additional models. The purpose of these will be to improve the collection of invoices and to assess the probability for a client to subscribe to the factoring offer.