How can AI be used to improve access to, management of, and utilization of public archives and documents?
The issue of modernizing access to administrative and heritage documents remains central to public policies on digital transformation. In this regard, ConversIA—an AI platform developed by the French startup DoXulting in partnership with the Tech.gouv program—marks a significant step forward in the automated management of public documents. Combining natural language processing, intelligent classification, and conversational dialogue, this solution aims to make government services more accessible, while raising new questions regarding governance, ethics, and transparency.
French AI to streamline document-related procedures
ConversIA is based on French language models specialized in understanding regulatory, legislative, and administrative documents. The goal is to provide a virtual assistant capable of automatically indexing diverse document collections (reports, circulars, digital archives), extracting key information from them, and responding to user queries—from both government agencies and citizens—in natural language.
This initiative comes at a time when the public sector generates tens of millions of pages of documents each year, many of which are difficult to use in their current form.¹ ConversIA offers a simple interface that allows users, for example, to ask, “What are the current regulations regarding the installation of solar panels in condominiums?” and receive a context-specific response, along with the relevant official documents1.
Use cases: from archives to government service counters
The first trials of ConversIA were conducted in local governments, with several practical use cases:
- Access to local council meetings: AI assistants enable both officials and citizens to instantly find meetings related to a specific keyword or time period.
- Easier access to regulatory texts: staff can use ConversIA to look up urban planning regulations or reporting requirements in their municipality.
- Leveraging heritage archives: In some museums and archives, AI is used to automatically extract metadata from thousands of scanned documents, speeding up their digitization.
Governance, ethics, and sovereignty: critical issues
The introduction of ConversIA into the French public sector naturally raises questions about data protection, algorithmic transparency, and the reliability of the responses it generates. To address these challenges, the solution relies on models hosted in France, with regular audits of the AI’s behavior and the ability to trace its sources and legal basis. It aligns with the recommendations of the Interministerial Committee on Public Transformation (CITP) regarding the responsible use of AI in government2.
In addition, ConversIA employs a systematic human oversight mechanism: each generated response includes a link to the source text, and administrative staff can annotate or correct the results, thereby enabling the system to continuously learn while ensuring accountability.
A National Strategy for Documentary AI
The development of ConversIA reflects a broader policy objective: to invest in publicly controlled, high-value-added digital infrastructure. The France 2030 program allocates 2.5 billion euros to AI, part of which is earmarked for vertical solutions for government agencies and local authorities3. The document management component is seen as a strategic lever, both for improving the relationship between citizens and the government and for strengthening the state’s informational sovereignty.
With ConversIA, France is thus demonstrating its ability to offer local alternatives to major cloud and AI platforms, while upholding its commitments to transparency and digital sovereignty.
What is the future of AI in public services?
Documentary AI paves the way for a more responsive, transparent, and inclusive government. But it also requires a rethinking of employee training, the validation processes for generated responses, and safeguards for neutrality. ConversIA thus lays the groundwork for an enhanced public service, but calls for constant vigilance regarding its uses, biases, and technological dependencies. Will citizens’ trust hold up in the face of this digital transformation?
References
1. Court of Auditors. (2023). Annual Report on Digital Public Archives.
https://www.ccomptes.fr/fr/publications
2. Interministerial Directorate for Digital Affairs. (2024). Guide to Responsible AI in Public Services.
https://www.numerique.gouv.fr
3. Ministry of the Economy. (2024). France 2030: AI Strategy for Public Services.
https://www.economie.gouv.fr/france2030