AI is an energy hog and you are selling more of it. What is the carbon footprint of YOUR platform — every scorecard you generate, every chat with your co-pilot?

The full executive answer

I will give you the honest answer, which has three parts: what we know structurally, what we measure, and what we have committed to. Structurally: the platform runs serverless — compute scales to zero when idle, no owned datacentres, no always-on fleet — with managed Postgres, and the heavy energy cost, AI inference, happens at Anthropic and OpenAI. Under the GHG Protocol our profile is essentially pure scope 3: purchased cloud and AI services.

What we measure today: honestly, not our own carbon — we have no measured baseline yet, and I will not perform precision I do not have. What we do measure is the precursor: every AI generation on the platform is metered — tokens, latency, cost, provider — through our own gateway. That consumption ledger is exactly the activity data a Software Carbon Intensity calculation needs; what is missing is applying supplier emissions factors, which is committed roadmap, not aspiration.

And here is the credibility anchor: the methodology is already shipped as product. Our green-AI module computes carbon as kilowatt-hours times grid intensity, treats grid intensity above four hundred grams per kilowatt-hour as a dirty-region flag, and recommends levers — right-sizing models, batching, region shifts. The commitment I will make in front of your panel: we run that module on ourselves and publish the footprint. Also worth saying plainly — a scorecard generation is a few seconds of inference, not a training run; the honest issue with boardroom AI is measurement discipline, not scale of burn, and discipline is the thing we ship.