AI is often presented as a shortcut to sustainability, but the reality is more complicated. The same technology that can help reduce waste and emissions can also increase demand, accelerate hardware turnover and create entirely new sources of environmental pressure.
Professor Jonatan Pinkse, Research Director, Centre for Sustainable Business at King’s Business School
04 June 2026
AI can help solve environmental problems, or make them worse
Researchers develop tool to make firms aware of how AI can reduce environmental harm yet also create new environmental risks, depending on how they manage and deploy it.
Artificial intelligence could help businesses cut waste, save energy and accelerate climate action, but new research from King’s Business School warns it could also deepen environmental harm if firms use it without clear sustainability guardrails.
It argues that AI is becoming the critical test of the “twin transition”, the idea that digital transformation and environmental sustainability must advance together.
The study, by Professor Jonatan Pinkse (King’s Business School), Professor René Bohnsack (Catolica Lisbon) and Professor Georg Reischauer (Leuphana University Lüneburg), challenges the assumption that adopting AI is automatically good for the planet. Instead, the researchers argue that AI’s environmental impact depends on why firms use it, how widely they scale it and whether they account for its hidden costs.
AI is already being used to forecast energy demand, improve smart grids, reduce food waste and optimise supply chains. But training and running AI systems can also use significant amounts of energy, water and hardware.
The findings also highlight a less visible risk. If AI makes a task cheaper, faster or easier, people and organisations may use it more often, reducing or even wiping out the original environmental benefit.
The research identifies three broad ways firms are likely to use AI. Some use it to amplify sustainability, by improving renewable energy systems or enabling new forms of environmental monitoring. Others use it mainly to stabilise productivity, such as automating office work, customer service or reporting.
The highest-risk cases are where AI amplifies harm. This can happen when the technology is used to expand fossil fuel extraction, drive greater resource use, extend the life of polluting assets or embed energy-intensive systems into sectors that previously had a relatively low computational footprint.
The researchers argue that firms should judge AI projects not only by return on investment, but also by whether they reduce environmental harm in absolute terms. Small efficiency gains should not be mistaken for genuine sustainability progress.
A firm can make one process more efficient while making the overall system more damaging. Businesses need to ask whether AI is reducing harm in absolute terms, or simply making unsustainable activity faster, cheaper and easier to scale.
Professor René Bohnsack, Catolica Lisbon
For policymakers, the research suggests that AI regulation and sustainability policy need to be considered together. Firms should be encouraged to disclose not only the benefits of AI-enabled efficiency, but also its environmental costs, rebound risks and enabled emissions.
AI has real potential to support climate and sustainability goals, but only if firms stop treating digital innovation as inherently green.
Professor Georg Reischauer, Leuphana University Lüneburg
The paper, Artificial Intelligence and the Twin Transition: The Good, the Bad, and the Ugly, is published in Industry and Innovation. It was authored by Professor Jonatan Pinkse, Professor René Bohnsack and Professor Georg Reischauer.
In this story
Research Director, Centre for Sustainable Business