In this paper, we ask what is the role of stakeholders such as citizens and the private sector in designing and implementing artificial intelligence projects in the public sector? Empirically, we use a comparative case study methodology focused on the experience of EmpatIA, a program led by the Latin American Open Data Initiative (ILDA). This analysis was done with privileged access to documents such as using financial and narrative progress reports, community calls, proposal slide decks as well as semi-structured interviews with each team. All the case studies in this paper have a common structure: objectives of the project, type of AI tool used, main beneficiaries, and evaluation of co-production dimensions. We use Nabatchi et al (2017) co-production typology to empirically analyse each case study. Our research finds that co-production in the field of artificial intelligence can occur at any level of co-production and phase of the service cycle, with the majority in the co-delivery of services. Practical lessons learned include the importance of data standardisation and the potential of co-production projects to contribute knowledge transfer from the private to the public sector