How Orano wins and delivers complex projects, one document at a time

In the nuclear sector, documentation is both the foundation and the bottleneck of operations. A project cannot be won before the RFP is answered, and cannot run without its safety files. Each deliverable combines text, tables, figures, cross-references, and regulatory data across dozens of pages.

The goal was not only to go faster but also to help improve structure and consistency across teams while maintaining a high level of expertise and traceability.

The solution: modular workflows

The first tests were carried out within support functions, where document formats were sufficiently standardized.

In just a few weeks, these teams were able to experiment with automated workflows for generating and summarizing documents that were previously drafted manually and were highly time-consuming.

Experiments quickly expanded to engineering teams, with agents capable of analyzing intervention reports, summarizing findings, and automatically completing technical grids based on multiple documents.

These more demanding use cases demonstrated that multi-agent systems could produce structured first drafts, offering engineers a reliable foundation for reviewing, modifying, enriching, and validating outputs.

‍

The results: time savings, ROI, and AI adoption

The benefits were concrete and measurable:

• Significant streamlining of documentation work
• Fast iteration cycles, with applications built and tested in just a few days

Documentation is no longer the bottleneck between a project won and a project delivered.

The next phase of the project is to continue exploring the potential of using Cobl for additional topics within Orano, and possibly turn it into a group-level “solution.”

The objective remains clear: use AI where it truly adds value, without seeking to automate everything, and without ever replacing expert validation.