With the support of DEEP, Generali Luxembourg has deployed an artificial intelligence solution on its own environment. This configuration enables it both to meet the constraints imposed by the supervisor in terms of preserving data confidentiality and to benefit from the possibilities offered by the large language models. The first use case, dedicated to AML/KYC operations, will considerably reduce response times to intermediaries, improve the quality of support and enhance the experience of employees.

"We quickly became aware of the technological shift represented by artificial intelligence." a game changer in terms of operability, comments Fabrice Walek, CEO of Generali Luxembourg. The life insurance company, which develops insurance and wealth structuring solutions, therefore looked at the technology with interest.

"The question that quickly emerged concerned the possibilities offered by the technology to gain in operational efficiency, service quality and user experience. With this in mind, two years ago we began an exploratory approach by setting up a dedicated team to identify use cases and define new pathways," continues Sébastien Rauth, Head of IT & Security at Generali Luxembourg.

The sovereignty of data and models is an essential prerequisite for any AI Solution in the Insurance field. In this context, we supported Generali in the deployment of a dedicated and disconnected AI, hosted in our Tier IV Datacentres in Luxembourg.
Nicolas Vivarelli

Nicolas Vivarelli, Head of Data & AI, DEEP by POST Group

Strong regulatory constraints, a framework that must be respected

When it comes to using artificial intelligence, players in the sector, under the supervision of the Commissariat aux Assurances, must meet a set of strict regulatory requirements, particularly in terms of data confidentiality. Insurance companies, in particular, cannot use the public cloud and, a fortiori, submit sensitive data to artificial intelligence models accessible in Software as a Service mode or to external agents. "The deployment of artificial intelligence therefore had to be considered in the light of these constraints. It was part of an ongoing project to renew our IT infrastructure, carried out in partnership with DEEP, with a view to boosting our performance as much as our operational resilience," continues Sébastien Rauth.

The insurance company has in fact equipped itself with a new high-performance system environment, managed directly by the teams of the POST group entity dedicated to professional digital needs, from one of its Luxembourg data centres. "Following on from this project, we have decided to deploy the artificial intelligence models we need to pursue our ambitions via DEEP, based on this sovereign infrastructure," adds Generali's head of IT.

A sovereign infrastructure serving AI

With its advanced expertise in data management and enhancement, DEEP has established itself as a first choice partner to support Generali in exploiting the possibilities offered by artificial intelligence. "Today, Generali benefits from a sovereign environment, entirely dedicated to the company, running carefully selected open source artificial intelligence models. The solution operates from one of our data centres in Luxembourg, with no connection to the outside world. This way, Generali can be sure that its data will not be exposed to third parties," explains Nicolas Vivarelli, Head of Data & AI at DEEP.

We identified artificial intelligence very early on as a potential lever for transforming our operations. A concrete 'game changer' that we needed to explore.
Fabrice Walek

Fabrice Walek, CEO, Generali Luxembourg

In order to set up this environment, necessary for the use cases envisaged, DEEP provided the Generali environment with GPUs - powerful processors that are essential for running the models - and supported the company from the selection of the large language models best suited to its needs, right through to the implementation of the solutions

The implementation of the solution was particularly rapid. It took just three months to deploy the first use case, from ordering the GPUs to interconnecting the AI with the insurer's information system, including setting up the dedicated environment and choosing and configuring the models.

AI serving AML/KYC operations

The solution, still in the POC phase, will be used to support the AML/KYC operations carried out by the company. "The artificial intelligence, as configured, analyses the documents provided by customers as part of an underwriting process, for example. This processing enables us to determine whether the data has probative value and whether it is consistent", continues Nicolas Vivarelli.

The technology saves considerable time in analysing information, as part of these operations required by anti-money laundering regulations. "These verification operations can account for between 40 and 50% of a customer's management time. Artificial intelligence acts here as an assistant, facilitating the processing of documents and the cross-checking of information. However, the final decision remains in the hands of a human operator," says Fabrice Walek.

Concrete gains for teams, partners and customers

The benefits associated with the use of artificial intelligence will be directly perceptible. "After a period of familiarisation, during which the prompts were fine-tuned to achieve a 98% satisfaction rate, processing times were considerably shortened. Where it used to take up to two days to respond to a new request, it will now take just thirty minutes. Customers will benefit from an improved experience. Staff will spend less time on tedious tasks with little added value, and will be able to handle a greater volume of transactions," comments Fabrice Walek. In short, we're saving time and improving the quality and relevance of the results. It's a win-win situation for everyone, which is essential for promoting acceptance of the technology."

Being able to deploy and operate our own AI models within a sovereign and controlled infrastructure is a key element of our strategy.
Sébastien Rauth

Sébastien Rauth, Head of IT & Security, Generali Luxembourg

A scalable solution, a responsible approach

In this respect, the choice of use cases is decisive. With this in mind, Generali has benefited from DEEP's expertise, both in deployment and in supporting teams in adopting the technology. "It is essential to ensure that the solution deployed is truly efficient and facilitates end-to-end processing. Otherwise, users will quickly turn their backs on it. The choice of the first use cases is therefore crucial. Our aim is to ensure, for each project, that we generate tangible added value for the customer," emphasises Nicolas Vivarelli.

The sovereign artificial intelligence solution implemented by DEEP is also scalable and enables new use cases to be envisaged, with the teams now looking to optimise the use of the power available, in particular by running other processing outside the AML/KYC analysis phases.

"The possibilities offered by this new environment are numerous. Our ambition is to continue this momentum, to explore the potential of the technology and to use AI to meet our challenges, through the choice of the most appropriate models, in order to maximise performance while optimising the resources used", concludes Sébastien Rauth. With this in mind, Generali Luxembourg can continue to rely on DEEP, which ensures responsible deployment of artificial intelligence by selecting the least energy-intensive models in particular, and by operating its client environments from Luxembourg data centres powered exclusively by green electricity.