Artificial intelligence is often presented as a magic wand, capable of streamlining administrations that are sometimes perceived as complex. Yet without a rigorous data management strategy, generative AI will remain little more than an impressive technological tool, incapable of scaling to an industrial level, says Ziad Choueiri, Programme Director in the Public Sector Business Unit at Sopra Steria.
The first challenge facing major public organisations seeking to implement AI solutions is the fragmentation and quality of their data. Too often, enthusiasm for the tool obscures the lack of substance. In the public sector, error is not an option. An algorithm trained on incomplete or biased data can have serious consequences for people's lives, whether in the allocation of social benefits or the calculation of taxes.
The mirage of AI without a data strategy
This is where the concepts of data quality and master data management (MDM) come in. Before deploying complex AI models, organisations must ensure the integrity, completeness and freshness of their digital data assets.
"AI is only as good as the data it consumes and is trained on," says Ziad Choueiri, Programme Director in the Public Sector Business Unit at Sopra Steria. "Today, the main obstacle is not a lack of algorithmic sophistication, but the quality of the information assets."
This behind-the-scenes work is how Sopra Steria enables major public sector players to turn proofs of concept into full-scale projects.
Siloed administration: The main barrier to holistic public AI
Today, public services are structured in tightly sealed silos. Whether it is the DGFIP, the CNAM or France Travail, each entity retains its own data in isolated repositories. This fragmentation is the primary obstacle to a public AI capable of delivering real value across administrations. "Siloed structures are the first glass ceiling to break. Without consolidated data, AI will remain a tool limited to the scope of the organisation it serves," says Choueiri.
Breaking down these silos is not merely an IT project; it is a democratic necessity. The goal is to create a 360-degree view of the user to deliver fairer services while ensuring that their data remains under control.
"For a long time, we believed that by bringing data back onto European soil, the sovereignty issue was solved. That is a mistaken belief. Tomorrow's true sovereignty will be decided at the next level, control of algorithms and overall data governance," he adds.
The quality of digital assets as the new foundation of the democratic pact
Beyond the technical dimension, the quality of digital assets is directly linked to issues of trust. At a time when disinformation is becoming industrialised, the state's ability to rely on reliable data is fundamental.
"Tomorrow's sovereignty will not be decreed through closure, but through the ability to produce and protect high-quality data that can be trusted," says Choueiri. This integrity legitimises public action and makes it possible to build innovative services without compromising values of fairness.
To be democratic also means rejecting the black box. Every decision made by an algorithm must be explainable and open to challenge by a citizen, explains Choueiri. "We must demand the supremacy of humans over algorithms: no algorithmic decision can be irreversible or made without the possibility of auditing and contesting it," he says.
Industrialising trust: The priority projects of the decade
The arrival of AI in the public sector cannot be a cold technological issue imposed from the top down. For Choueiri, it is crucial to bring public servants on board and to promote new skills. "AI must not be perceived as a replacement, but as an assistant capable of relieving staff of repetitive tasks so they can refocus on human support," he says. Education and communication are key to addressing fears linked to large-scale automation.
The 2030 horizon sketches an ambitious digital Europe that fully leverages AI's tremendous potential. But this future will not be built on wishful thinking. It requires a European political will to make massive investments in this disruptive technology. But the benefits are huge. "If we make the right choices now, by 2030 we will have European champions at every level of the AI technology stack," concludes Choueiri.
By investing now in data governance, software and sovereign infrastructure, Europe can build the sovereign AI needed to support public action by 2030 and beyond.