University of Manchester researchers have built an AI planning workflow that helps UK councils, hospitals and schools decide which building retrofits to fund first. It combines whole-life cost, carbon impact and service disruption into one scoring system, with stakeholder priorities turned into transparent weights. The system uses reinforcement learning to search retrofit sequences and a Bayesian neural network to estimate energy, carbon and cost outcomes with confidence ranges. In tests on 1000 building and budget scenarios, the AI planner ran about 30 times faster than enumerating every feasible sequence. It also produced better ESG results than common baselines like lowest whole-life cost, carbon per pound ranking, or static multi-criteria analysis. The authors stress the tool is not a black box. Every choice can be traced back to the weights and constraints set by the public body using it. The current evidence comes from public datasets and scenario analysis, not live portfolios. Next steps are metered data, digital twins and real deployment. If those trials hold up, the workflow could become a practical way to spend public decarbonisation budgets with less guesswork and stronger audit trails.
