Effective management of marine ecosystems requires understanding complex species interactions and food web dynamics. We present two artificial intelligence approaches that help tackle this challenge. The first uses generative AI to develop Models of Intermediate Complexity for Ecosystem Assesment (MICE) (Plagányi et al 2014) that can capture key ecological processes while avoiding unnecessary ecological complexity. We demonstrate that given effective prompting, generative AI systems can build, parameterise and calibrate Crown-of-Thorns starfish populations on the Great Barrier Reef. Our second framework addresses a common challenge in marine ecology: how to parameterise diet matrices for marine ecosystem models like Ecopath with Ecosim (EwE) (Guesnet et al. 2015) and Atlantis (Pethybridge et al 2019). We developed an approach to use generative AI to consistently group species into functional groups and estimate diet proportions between groups. We tested this approach across three diverse Australian marine ecosystems, and show that the resulting diet matrices can provide a suitable starting point for calibrating these marine ecosystem models. These frameworks help bridge the gap between ecological theory and practical management by automating complex analyses while ensuring ecological realism. This enables more widespread adoption of ecosystem-based management approaches across different marine environments, from coral reefs to temperate systems.