AI in local government: public services move to action

After two years of experiments, French local authorities are industrialising their first AI use cases: pre-processing of planning applications, citizen-response assistants, energy optimisation of building stock, prioritisation of road maintenance. The movement now reaches mid-size cities and inter-municipal bodies, no longer just metropolises.

The framework has clarified: the AI Act classifies most of these uses as limited risk, the CNIL has published its reference frameworks, and public procurement now has standard clauses for buying AI solutions. Perceived legal obstacles are now largely solvable by method.

Observed success factors: start with high-volume, low-legal-complexity processes (civil status requests, appointment booking, recurring questions), always keep a human access channel, and train staff before deployment rather than after.

Cardan-AI recommendation: for a local authority, the right starting point is a diagnostic of citizen-request flows. It almost always reveals two or three areas where AI improves both response time and staff working conditions.