Carbon markets are relatively nascent. But given how new they are compared to more established markets, they're also subject to meaningful overlapping regulations and standards. That means supply can be either more or less elastic to demand when you look at it on a granular level. What do we mean? Supply forecasts that look at the totality of the VCM look significantly different when you project only CCP-labeled supply. And when you layer on an individual buyer's spec, it changes the picture even more. The trouble is those sorts of forecasts require deep understanding of the CCPs, the rate that ICVCM is labeling projects, and the rate those projects will issue credits. The only way you can do it accurately is through project-level analysis of the entire VCM. It's not impossible to do, but it requires human expertise in addition to the raw data itself. And even then, it's only economical with the help of purpose-built AI software.
Many technology companies are calling themselves AI-native service firms this year. Here's a test: can you point to work you delivered for one client, in their specific context, that no expert team could have produced at that speed or that price? At Patch we do, every day. A recent example. A client asked how much CCP-eligible carbon credit supply would exist that actually meets their procurement spec. Answering it meant assembling ICVCM review status, methodology and registry data, project documents, forward issuance estimates, historical approval and issuance behavior, and explicit assumptions about how the market evolves between now and 2030. That data is fragmented, inconsistent, and buried in documents running hundreds of pages. It also can't be answered by counting what exists today. In its May 2026 decisions, the The Integrity Council for the Voluntary Carbon Market (ICVCM) counted 6.44 million credits issued under ACM0008, the coal mine methane methodology, and few of them are expected to clear the CCP label once the approval conditions apply. A current count tells a buyer what exists under today's rules. It says almost nothing about what will be issued, approved, rated, and available to buy in 2030. AI handled the breadth: document review, extraction, normalization, and analysis across a corpus no team would ever read line by line. Our experts handled the judgment: which sources are authoritative, which variables actually bind the client's spec, which assumptions are defensible, and how to present a result that stays uncertain. A traditional consulting team could have done this work. The cost and the timeline would have capped the client at one or two questions when they had a dozen. A generic AI tool could have produced an answer too, with no market context behind it and nobody accountable for it being right. That is what AI-native services means in practice. The economics change, so clients stop rationing the questions they ask. Full analysis in the comments.