Show me the green-AI module properly. My organisation is deploying AI everywhere — how would I use this to govern the carbon cost of that estate?

The full executive answer

The module treats every AI workload in your estate as a governed asset. You register each one — classified as training, inference, or data pipeline — with its region, monthly energy draw, the grid carbon intensity for that region, and optionally its monthly cost. From that it computes monthly CO2-equivalent deterministically — kilowatt-hours times grams per kilowatt-hour — and rolls the estate up into a footprint view. Deliberately simple, auditable math: no black box between your data and your number, which matters if the number ever feeds a disclosure.

The governance value is in the levers. Any workload sitting on a grid above four hundred grams of CO2 per kilowatt-hour gets flagged as dirty-region, and the module recommends efficiency levers matched to the workload type and its optimisation status — model right-sizing and batching for inference, scheduling and region-shifting for training, with each workload tracked from unoptimised through optimising to optimised. So your AI carbon posture becomes a managed backlog with statuses, not a one-off report.

Where it fits your stack: it is the strategy-and-governance layer, not a metering agent — the energy inputs come from your teams or your cloud billing data. And it connects to the rest of the platform: the footprint story feeds the ESG assessment's environmental pillar and the net-zero pathway, so AI carbon shows up inside board-level ESG strategy rather than in a spreadsheet nobody reads. Pair it with the ESG assurance registry and the AI-estate footprint becomes a disclosure line with an owner and a deadline.