Share with your COO
The case for applying AI to corporate sustainability programs is real, but the article’s most valuable contribution is its skepticism about where that value actually lands. Gartner’s Jackie Swanson frames the core test clearly: measurable physical output, trustworthy data, and a decision someone will act on. Against that standard, most GenAI-powered ESG reporting tools fail on at least one count. The AI tools for sustainability professionals survey covers eight enterprise platforms, from Watershed to IBM Envizi, while cataloguing where classical ML beats frontier models on both cost and carbon.
What this means for your business
The organizations most exposed here are the ones that bought sustainability software primarily to produce board-ready ESG disclosures. If the underlying data is inconsistent across suppliers and procurement systems, adding an LLM layer doesn’t fix the data problem, it just formats the noise more attractively. COOs who own the operational systems that generate that data sit closer to the actual fix than the sustainability team running the reporting tool on top of it.
The useful analytical split in this piece is between AI applied to physical systems and AI applied to documents. Predictive maintenance on industrial equipment, energy grid optimization, logistics routing, these produce kilowatt-hours saved or downtime avoided, numbers that show up in operating budgets and are auditable. ESG reporting automation produces summaries, and as Swanson puts it, a model pointed at untrustworthy data returns “confident, beautifully formatted estimates.” That distinction should change how you evaluate vendor demos: ask which category the product is actually in, not which one the pitch deck claims.
The accountability gap Gasilov identifies is the sharpest structural point in the piece. IT enables access, finance pays licensing, facilities absorbs the electricity bill, and sustainability reports the emissions. Nobody fully owns AI’s environmental cost, which means nobody is positioned to weigh it against AI’s environmental benefit. That’s not a governance abstraction, it’s the specific condition under which organizations end up running frontier LLMs for tasks a smaller model would handle fine. The COO is the one executive whose scope spans all four of those cost centers, which makes this genuinely their call to own.
Swanson’s closing metric is the one worth pinning to your next AI portfolio review: measure whether anyone acts on the model’s output. A model that generates sustainability insights nobody uses is pure cost, and the carbon it consumed is unambiguously additive to the emissions you’re trying to report down. If your current AI sustainability stack can’t answer that question with usage data, that’s the renewal conversation to have before the next contract cycle, not after.
Concept deep-dive: Scope 3 emissions
Scope 3 covers all the greenhouse gas emissions a company is indirectly responsible for across its value chain, suppliers shipping raw materials in, customers disposing of finished products out, everything except what happens inside the company’s own walls or power meters. It’s the hardest category to measure because the data lives in thousands of external systems the company doesn’t control. That’s precisely why AI-assisted data collection looks attractive here, and precisely where data quality failures do the most damage to ESG credibility.
Based on reporting from Most promising AI tools for sustainability professionals, originally published 2026-07-30 06:15:00.
