City of Helsinki's AI Experimentation Accelerator — Rapid, Low-Budget AI Sprints Across Municipal Services

Modelled on an internal accelerator used by the Finnish Broadcasting Company (Yle), the City of Helsinki sought a way for municipal staff to test AI and software-robotics ideas drawn from their own daily work, without resorting to large, centrally procured IT programmes.

The Experimentation Accelerator aims to let city staff propose and test AI and software-robotics use cases with small dedicated budgets in recurring three-month sprints, offering a replicable low-cost, staff-driven alternative to large centrally procured AI programmes.

The programme runs recurring three-month sprints in which staff propose experiments drawn from their daily work. Examples include AI support for jobseeker services, shift planning for social- and healthcare employees, an AI-powered personalised sightseeing guide, machine-learning analysis of aerial photographs to track the physical deterioration of pedestrian crossings over time (with the code published openly on GitHub), workforce forecasting for library services, and improved information management in infrastructure construction. A smaller set of experiments — including a city-events recommendation algorithm and a resident-feedback and participatory-budgeting text-analytics tool — were selected to move toward wider use.

The city reports around 80 experiment proposals submitted by staff, of which 35 reached the live experiment phase. A smaller subset of experiments was selected to move toward wider use.

The programme has been documented as a replicable low-cost, staff-driven model by the ITU's AI for Good initiative and industry case studies, positioning it as an alternative to large, centrally procured AI programmes; most individual experiments, however, remain at sprint/pilot scale rather than city-wide deployment.