Artificial intelligence (AI) is impacting all aspects of food systems, including production, food processing, distribution, and consumption. AI, if implemented ethically for sustainability, can enhance biodiversity, conserve water and energy resources, provide land-related services, power smart cities, and help mitigate climate change. However, there are significant issues in using AI to transition to sustainable food systems. AI’s own carbon footprint could cancel out any sustainability benefits that it creates. Additionally, the technology could further entrench inequalities between and within countries, and bias against minorities or less powerful groups. This paper draws on findings from a study of the Victorian Farmers’ Markets Association (VFMA) that investigated the complexity of designing AI tools to enhance sustainability and resilience for the benefit of the organisation and its members. Codesign workshops, both synchronous and asynchronous, semi-structured interviews, and design innovation methods led the VFMA to experiment with an AI tool to link sustainable soil practices, nutrient rich produce, and human health. The analysis shows that the codesign process and an agile approach created a co-learning environment where sustainability and ethical questions could be considered iteratively within transdisciplinary engagement. The bottom-up approach developed through this study supports organisations who want to engage with AI while reinforcing fairness, transparency, and sustainability.