Four assignments, measured results.

Nexans

Framing then product leadership for data and AI products at a listed industrial group

The group's Digital and Data department wanted to move from scattered generative AI experiments to products held in production. The assignment began with framing: qualifying use cases, a roadmap shared between business and technical teams, and setting up the governance that structures exchanges between these stakeholders.

It continued as product leadership on the group's data platform, from design to production: identifying and mapping source data, functional and technical specifications, coordinating development, acceptance testing, then supporting internal teams as they took ownership.

The team was strengthened with a software architect profile to speed up production releases and extend the business scope covered.

Sponsor: Digital and Data department. Scope: use-case framing, data governance, product leadership, production release. Ongoing without interruption since late 2023.

Air France-KLM

Leading the automation and artificial intelligence theme for customer service technology

The group's customer service is operated from around thirty centres worldwide. Several teams had launched automation and artificial intelligence initiatives there, each starting from the same initial questions.

The assignment consisted of defining the automation and AI strategy for the customer service technology department: a review of the existing situation, identifying and prioritising use cases, and building the associated value case.

It then covered coordinating current and upcoming initiatives, maintaining a consolidated view of use cases, and running the internal community so teams would stop addressing the same problems separately.

Sponsor: customer service technology department. Scope: strategy, use-case prioritisation, value case, roadmap, coordination of initiatives. Customer service operated from around thirty centres worldwide. 2024.

Axylis

Data and artificial intelligence diagnosis for an accountancy, advisory and audit firm

The accountancy profession is at a turning point: e-invoicing will transform the core of the business, and a significant share of historical activity is set to shrink. The firm wanted to know what artificial intelligence could genuinely bring, and in what order, before committing to anything.

The diagnosis began on the ground: fifteen interviews, twenty-five people met across every level of the firm, and sixty-six use cases identified. This raw material was then qualified against three criteria: commercial value, technical feasibility, and the cultural change required.

The work concluded with a discussion with the executive committee to confirm the vision and objectives, followed by a presentation and a prioritised roadmap. The firm then had a basis to decide and fund, instead of a list of intentions.

Sponsor: firm management. Scope: data and AI diagnosis, identification and prioritisation of use cases, roadmap. 15 interviews, 25 people met, 66 use cases identified.

Chamber of Notaries of the Caen Court of Appeal

A documentary assistant that always cites its sources

The Chamber faces a growing number of notarial offices and a doubling of its notary population. Questions pour in on professional ethics, office management, accounting and employment law. That expertise sits with a handful of people and is scattered across hundreds of documents.

We framed and led the build of an assistant based on retrieval-augmented generation coupled with a knowledge graph. The corpus went from 120 to more than 900 searchable documents, including thirty-five years of ethics opinions. Every answer cites its sources, and an expert notary validates continuously.

Timeline: three months of framing and document sourcing, three months of prototype, three months of iterative acceptance testing with the business expert, three months of industrialisation. In production since October 2025. Hosted in France, GDPR and EU AI Act compliant. Roll-out to the national network under way.

+76% accuracyAverage score up from 2.72 to 4.79 out of 5 between the first and second evaluation campaigns.

Use case published on the Bpifrance Accélérez avec l'IA platform