We are now at the stage where digital leaders have some experience in how to begin the cultural adoption of Artificial Intelligence (AI). At major bank BNP Paribas Fortis, Chief Data Scientist Manuel Piette heads up the AI Center of Excellence (CoE) at the Belgian arm of the bank. Piette is using communities, Domino data technology, and Europe’s frontier AI technology Mistral to improve data management and speed the adoption and usage of AI. Piette was in London and shared his CoE approach with us.

BNP Paribas Fortis was created in spring 2009 following the acquisition of Fortis Bank in Belgium by BNP Paribas. It is the largest retail bank in Belgium, offering a full range of services to retail customers, as well as business banking to both small firms and enterprises.

Piette has been with the organization for 21 years in a variety of data analytics roles supporting marketing, retail and private banking, and now is the bank’s Chief Data Scientist and Head of AI. A centre of excellence has been developed by Piette to help teams across the bank learn and adopt AI, especially as his team has developed and deployed the BNP Paribas Fortis generative AI platform, a secure large language model (LLM) for the 11,000 employees. He describes the approach of a CoE to AI as:

We focus on four main areas: deployment of AI, improving the customer experience, improving employee productivity and the optimization and automation of processes, such as fighting fraud and customer protection.

On employee productivity, Piette says he counters fears about AI with staff by asking them to use their imaginations:

They can find new ways to do their old job. I tell people just to try it. For me it is about being creative. We bring generative AI to the employees so they can express their creativity and increase their productivity.

By fostering creativity, Piette aims to help both the employees and the bank with the changing skills needs that AI is bringing about. He adds:

Generative AI has brought a need for new skills. We provide training for our data scientists, but we also want to attract new talent that has already used AI in other environments.

Whether a data leader or digital and technology leader, it is a balancing act we increasingly hear from our community: a pressing need to retain talent that really understands the business, its customers and where its operations are a unique competitive advantage, but also a realization that there are potential employees that have imagined using AI in ways that could change business operations for the firm. Piette says this transition is part of the normal narrative of data leadership and its hiring needs; he says:

20 years ago, we needed engineers and statisticians, then the market evolved, and we needed people that could implement data technologies and integrate them with our environment. Now we are moving into the era of agentic AI, so we still need people with software engineering skills, but also machine learning skills.

The CoE is helping employees by carrying out what Piette describes as journaling sessions every two weeks. At these, the data scientists in his team prepare a learning track on subjects like how to evaluate an AI agent, agent integration, making changes to an AI bot, voice native AI technologies, as well as how the staff of BNP Paribas Fortis can utilize data technologies such as Domino Data Lab. He adds:

We invite our partners in to open the minds of the community and help them learn.

Like many data leaders, managing and extracting value from unstructured data is a challenge for Piette, and he believes AI has potential. He describes unstructured data as the unseen half of an iceberg and a real challenge for the bank. The bank is deploying AI to teams that manage large and growing volumes of structured and unstructured data, stating that AI is:

Used to enhance internal document search, automate data extraction and support complex financial analyses requiring cross‑referencing multiple sources.

Open source, European

Piette’s team has two key technologies at hand: the Domino Data Lab data management platform and Mistral, the French-headquartered frontier AI model. The data scientist says Domino has been instrumental in “harmonizing” the data of BNP Paribas Fortis so that it can be used across all teams and business units.

The bank recently renewed its partnership with Mistral AI, a partnership that began in 2023 and will continue for another three years. BNP Paribas and Mistral will work together on research projects, and the bank will continue to use the Mistral LLM and software. Announcing the extension, the bank said the deal will focus on high-value use cases in corporate and institutional banking, and commercial banking.  

For Piette, Mistral being open source is really important, as is the ability to bring the models in-house and therefore protect customer data and ensure data sovereignty. In a statement, BNP Paribas said:

The partnership also reflects BNP Paribas’ choice to work with a European technology partner on selected use cases, as part of a broader, multi-model approach to generative AI, guided by performance, data sensitivity and geographic considerations.

That includes know your customer (KYC) processes, which have high levels of data privacy attached to them and a growing list of compliance requirements. Despite this, the bank said Mistral will help:

Automate low value-added steps and orchestrate workflows more efficiently, while maintaining full expert validation of risks and decisions. BNP Paribas and Mistral AI are also collaborating on selected research projects and specialized models, aligned with specific business and risk management needs.” 

My take

The number of data and digital leaders coming forward to explain how they are investing in training to benefit both the bottom line and the talent of the organization is reassuring. Piette’s work to involve his suppliers like Domino Data Lab ensures recipients get a wide view of the potential and risks of AI.

The partnership between BNP Paribas and Mistral is an important one to watch, as data sovereignty threats keep increasing; having a non-US or Chinese AI player in the pack is going to be increasingly important.