14 October 2026 Notes from a seminar on 14 October 2026 on Real-time carbon intelligence: turning digital impact into better decisions. Part of a series of events on responsible AI by ClimateAction.tech with ustwo (previously: AI sweat). Any errors are mine. (answering the question in my headline:) Maybe but there’s a long way to go “Every digital decision has a physical consequence. The AI model we choose, the cloud region we use and the way a product scales can all change the energy, carbon and infrastructure behind it, yet most teams cannot see those consequences while they are building.” From activity -> report -> learn … to activity -> information -> decision. Environmental info can become more than a post facto accounting/documenting effort. [Also suggests there’s no (or not enough) visibility at the point of decision.] James Hall GreenPixie: GP are a “Greenops” data provider - maintain a coefficient db of 11m+ cloud “items”. Looks interesting. C suite say save energy/cut costs but teams are making decisions that affect those outcomes, but without data or visibility on how their decisions affect the wider goals. energy/carbon/water a bigger driver of decisions than cost? [can this be true?] Artificial Analysis (company?) done loads on assessing tokens Greenpixie data explorer and methodology [will be interesting to compare to the recent Cleer data. Obv we’re in an interesting time: for many orgs, data/AI emissions are both very small compared to their emissions from heating, travel and industrial processes, but at the same time their data/AI use is growing really quickly.] training course Greenpixie: resource use across model and physical infra Greenpixie: green metrics and risk metrics Mark Buss Senior Digital Sustainability Manager OVO Energy, co-organiser [Green Software Brighton]:(https://greensoftwarebrighton.co.uk/) Data from cloud providers etc at least 2 week lag? Working with Greenpixie to fill their gaps measuring data across their pods (groups of teams) OVO: Cloud provider data can drive discussion on areas to improve or investigateOVO: team carbon intensity: some teams are large, some teams use a lot of AI Paolo Rizzi ustwo: Prism estimates impact of AI activity and presents it inside the developer’s existing workflow. Improve understanding over time? Before Prism: 43% trying to reduce unnecessary AI use, 36% making no effort. After Prism: 82% and 9% Awareness easier to shift than confidence (and therefore action?): Prism made them more aware of carbon but didn’t make them more confident taking action Some of the most consequential decisions happen [strategy? selecting tools?] before any is made about code ustwo Prism: Visibility influenced behaviour but awareness is easier to shift than confidence to act Questions: AI enabling vs reducing emissions [some of this is pure Jevons paradox], incentives for using AI, Big AI as bad actors Is there an equivalent of a “B Corp” for ai foundation models? I guess it is B Corp itself. [People wanting more reponsible AI should perhaps hope for an AI capex/financing crash that resets the debate away from speed and capacity to responsible resource use.] Tokenomics foundation Carbon budgets - flat fee subscription models create incentives that sometimes point in opposite direction to emissions Greenpixie: Tokenomics