Cities often have either long lists of proposed actions or no actions at all. In any case they have no way to evaluate which actions will deliver the most climate benefit. Staff need to balance emissions reductions, adaptation, cost, and local constraints. Without support, decisions are political or arbitrary.

Existing prioritization tools are complex, not built for cities, or too rigid. Many cities end up picking actions without understanding their impact or fit. For making an informed decision, they need to process huge amounts of systemic and interconnected variables

We knew that for a problem like data processing constraints, AI is the solution. At the same time we had expertise on greenhouse gasses inventories, which we could use as input. We designed a flexible tool that adapts to each city's context and priorities. Our approach focused on transparency and speed: let cities input what they know, get rankings, and adjust assumptions on the fly.

We mapped common prioritization frameworks and developed a ranking machine learning model, complemented with expert input and LLM-powered climate implementation plan generation. I helped to design and enhance an interface to show real-time changes in priority rankings and allow the user to customize it to generate multiple action plans. Everything was tested in real decision-making sessions with city teams.