According to PwC, GenAI could improve operational efficiency, which might indirectly reduce carbon footprints in business processes.
Companies can implement strategies to reduce energy consumption during the development, customization, and inference stages of AI models. By leveraging GenAI applications, companies can offset emissions in other areas of their operations.
Discover sustainability AI applications with real-world examples that leverage AI to build a smarter, more efficient, and more sustainable future.
How is the sustainability of AI being evaluated
As the use of AI expands across sustainability initiatives, questions are increasingly being raised about how the sustainability of AI itself should be assessed.
Recent research and policy discussions suggest that improvements in efficiency or emissions reductions alone are insufficient to evaluate long-term impact. A broader evaluation is needed to understand the environmental, social, and structural consequences of developing and deploying AI systems.
Here are some perspectives from the Sustainable AI Conference in September 2025used to assess whether AI applications meaningfully support sustainability goals beyond short-term operational gains.
The key takeaways from the conference are that AI can only be considered sustainable if it addresses environmental, social, political, and justice impacts together, because unlimited scaling and efficiency-only approaches risk reinforcing inequality, extractivism, and structural harm despite technical gains.
Sustainability goes beyond energy efficiency
According to the conference, sustainability is a broad concept rather than a narrow technical metric. Many contributions argue that focusing only on energy efficiency or carbon reduction misses key impacts of AI systems.
Sustainability should be discussed across multiple dimensions:
- Environmental costs such as energy use, water consumption, minerals, and e-waste
- Social effects, including labor conditions, inequality, and gender impacts
- Political and economic issues, such as power concentration and control over infrastructure
- Knowledge-related concerns like loss of epistemic diversity and weakened critical thinking
The overall position is that AI cannot be considered sustainable if it performs well environmentally but causes social or structural harm.
Scaling AI conflicts with sustainability goals
A recurring theme is the tension between large-scale AI development and sustainability. Current AI trajectories emphasize bigger models, more data, and higher compute demands, while sustainability requires limits and selectivity. See LLM scaling laws for more.
Several researchers highlight alternative directions:
- Smaller, task-specific models instead of general-purpose systems
- Local or domain-bound deployment rather than global scaling
- Careful justification for high-performance computing use
- Clear distinction between essential and non-essential AI applications
The argument is not that scaling is always wrong, but that unlimited scaling is incompatible with long-term environmental and social constraints.
Power and extractivism are central concerns
Many contributions frame AI sustainability as a question of power rather than technology alone. AI systems depend on global supply chains that often rely on extractive practices.
Key issues discussed include:
- Data extraction from marginalized and Indigenous communities
- Resource mining justified by green transition narratives
- Concentration of compute, cloud services, and data centers in a few regions
- Corporate control over energy infrastructure linked to AI deployment
From this perspective, sustainability claims are weak if they ignore how benefits and burdens are distributed across regions and populations.
Justice-based frameworks dominate the discussion
Justice is treated as a core requirement for sustainable AI. Several ethical lenses are repeatedly applied to assess AI systems.
Common frameworks include:
- Energy justice, focusing on who pays energy costs and who benefits
- Feminist ethics, emphasizing care, recognition, and relational impacts
- Decolonial and Indigenous approaches, highlighting data sovereignty and consent
- Structural responsibility, which looks beyond individual developers to systems and institutions
Across these perspectives, a shared conclusion emerges: AI that reinforces inequality or oppression cannot be considered sustainable.
Governance mechanisms are insufficient
Legal and policy-focused papers argue that existing governance frameworks lag behind the material realities of AI systems. Environmental impacts are often weakly regulated or treated as voluntary concerns.
Identified gaps include:
- Limited requirements to measure and disclose AI environmental impacts
- Weak enforcement mechanisms in existing AI regulation
- Overreliance on corporate self-reporting
- Difficulty applying individual rights frameworks to structural harms
Alternative AI pathways are proposed
Despite criticism, the conference does not reject AI altogether. Many contributions outline alternative ways to develop and use AI that align more closely with sustainability.
Proposed directions include:
- Small and efficient models designed for specific contexts
- Public-interest and open-source AI infrastructures
- Participatory and community-led AI design processes
- Degrowth-oriented approaches that prioritize sufficiency over expansion
AI agents in sustainability
AI agents in sustainability are autonomous or semi-autonomous systems that use artificial intelligence to perform specific tasks related to environmental, social, and governance (ESG) goals.
They analyze sustainability data, identify trends, and execute actions with minimal human input. These agents combine data processing, natural language understanding, and machine learning to support decision-making and operational efficiency in sustainability management.
Their primary purpose is to reduce the manual work required to gather, analyze, and report sustainability data. By automating repetitive, data-intensive tasks, AI agents enable sustainability professionals to focus on strategic planning, compliance, and performance improvement.
Depending on their level of autonomy, they can either work independently or assist human teams in completing defined processes.
There are generally two types of AI agents in sustainability:
- Autonomous agents: These function independently, making data-driven decisions and executing actions without direct human supervision.
- Assistive agents: These support human teams by offering recommendations, analysis, and automation for specific tasks.
Real-life example: CO2 AIautomates carbon management and converts sustainability commitments into measurable outcomes. The platform reduces repetitive, data-intensive tasks, allowing sustainability teams to focus on analysis and emissions reduction.
Its AI agents address issues such as inconsistent data, complex carbon calculations, and supplier engagement by automating data cleaning, standardization, and emissions estimation at scale.
The system also supports compliance with frameworks and regulations, including SBTi, CSRD, CBAM, and SB253, while ensuring data security and regional data control.
Data Agent
- Standardizes data from multiple sources within minutes.
- Structures large datasets into audit-grade, compliant formats.
- Enables accurate and transparent emissions reporting.
Scope 3 Agent
- Identifies and retrieves verified supplier emissions data.
- Recognizes and matches supplier entities using the company and purchasing context.
- Assesses supplier maturity based on reporting quality and target commitments.
Emission Factor Matching Agent (EFM Agent)
- Matches products and materials with the most relevant emission factors across extensive databases.
- Performs semantic analysis to interpret technical terms and ensure accurate matches.
- Allows large-scale emissions estimation at a fraction of the cost of traditional life-cycle assessment.
1. Data and reporting automation agents
AI agents are frequently used to collect, verify, and structure sustainability data from multiple internal and external sources. They can process large datasets to ensure data integrity and compliance with reporting standards.
- Automating ESG and sustainability reports according to frameworks such as ESRS, SASB, CDP, and GRI.
- Preparing sections for regulatory filings, such as 10-K reports, and maintaining audit trails.
- Aggregating emissions data, resource usage metrics, and other key indicators for consistent analysis.
2. Stakeholder engagement and communication
AI agents assist in managing communication with internal and external stakeholders who require sustainability data or updates.
- Answering investor or regulator queries using verified data.
- Automating supplier questionnaires and sustainability surveys.
- Generating tailored sustainability summaries for executives, customers, or the public.
3. Operational efficiency and resource management
AI agents use predictive and optimization models to improve sustainability-related operations.
- Monitoring equipment and predicting maintenance needs to prevent waste and downtime.
- Evaluating supplier performance to support sustainable procurement decisions.
- Optimizing logistics and field operations to minimize emissions and resource use.
Preparedness for natural disasters
AI plays a crucial role in improving disaster preparedness and response, especially as extreme weather events become more frequent and intense due to climate change. By powering systems such as flood-warning networks and forest-fire detection, AI helps save lives, protect livelihoods, and minimize economic losses.
Using real-time data, predictive models, and advanced sensors, these technologies provide early warnings and actionable insights, giving us a better chance to reduce the devastating effects of natural disasters.
Real-life example: Google has introduced Google Earth AI, a suite of geospatial AI models and datasets designed to address global challenges such as weather prediction, flood forecasting, and wildfire detection.
Part of this launch includes AlphaEarth Foundations. These models support urban planning and public health by analyzing imagery and population data, and they already power features in Google Search, Maps, Earth, and Cloud.
The goal is to provide actionable insights to help solve critical environmental and societal issues.
AlphaEarth Foundations is an artificial intelligence model designed to integrate extensive Earth observation data into coherent digital representations of terrestrial land and coastal waters.
By processing petabytes of satellite imagery and geospatial measurements, the model enables more accurate and efficient environmental monitoring and analysis. Its outputs, referred to as embeddings, are now available through the Satellite Embedding dataset on Google Earth Engine.
These embeddings have already supported over 50 organizations, including the United Nations and academic institutions, in efforts such as ecosystem classification, agricultural assessment, and land use monitoring.
AlphaEarth Foundations also demonstrates significant improvements in data compression and mapping accuracy, and can serve as a foundation for future integration with general-purpose AI systems.
4. Flood warning
According to recent data, 250 million people are affected by flooding yearly. PwC suggests that AI-driven improvements in flood warning systems could save more than 3,000 lives and reduce economic damages by up to $14 million. These technologies provide timely alerts, helping communities take action before disaster strikes.
Real-life example: Google’s operational flood‑forecasting system, based on a large LSTM-based language model for hydrology, was launched in 2018. It combines two AI models: a hydrologic stage‑forecasting LSTM that predicts river levels, and an inundation model (using threshold and “manifold” algorithms) that simulates the flood extent and depth to generate alerts up to seven days in advance.
The system currently covers over 100 countries via “virtual gauges” and verified river basins, reaching approximately 700 million people with flood-forecasting alerts delivered through Google Search, Maps, Android, the Flood Hub, and government partners.
Key achievements include:
- Flood forecasting via LSTM stage and inundation models.
- Mature deployment since 2018 in over 100 countries.
- Up to a 7-day lead time with real‑time alerts to 700 million people.
- Strong evidence through Nature/HESS publications.
Figure 1: The image illustrates Flood Hub’s global reach, showing how it supports flood forecasting for more than 700 million people.
5. Forest fires
AI is also a powerful tool in the fight against forest fires, helping to prevent devastating losses. Drones, satellites, and sensors on tall towers continuously monitor forests, detecting signs that could indicate the start of a fire, such as unusual hot spots or rising smoke.
With proper training, AI can learn to differentiate between smoke and other forest scents, providing more accurate early warnings and quicker responses to potential fires..
Real-life example: Dryad Networks has installed around 400 “electronic noses” in the Eberswalde forest in Brandenburg, a region heavily affected by wildfires. These devices can detect gases during the earliest stages of a fire while also monitoring temperature, humidity, and air pressure.
By providing real-time data on environmental conditions, these sensors help identify potential fire risks early on, improving the ability to respond quickly and minimize damage..
Fighting air pollution
Air pollution is getting worse, and it can escalate to a global public health and environmental emergency that causes over seven million premature deaths every year and $8.1 trillion in health damages alone.
AI can help reduce air pollution with real-time warnings and predictive models:
6. Real-time warnings
With data provided from air quality monitors, AI can offer insights on the impact of air quality on people and help decide on health protection policies.
Also, by processing data from different monitors in real time, it can send out alerts when pollution levels spike. This way, people can act right away: stay in or wear masks.
Real-life example: IQAir application has a ranking that shows in real time which towns have the most pollution in the air. The Plume Labs app gives full maps that show where pollution is worst. The app also tells what the air quality will be like every hour, as levels can change significantly during the day.
7. Predictive models
Developed with AI and machine learning models, predictive models can forecast information such as air-pollutant concentration.
Real-life example: Engineers from Cornell designed a model that can calculate the fine particulate matter (PM2.5), which is the soot, dust, and exhaust from trucks and cars that gets into people’s lungs. Using these models, future negative impacts can be avoided.
Biodiversity
8. Biodiversity monitoring and conservation
Conserving biodiversity is one of the biggest challenges brought by climate change. AI offers solutions for improving biodiversity monitoring and conservation.
Technologies such as neural networks, computer vision, and satellite vision can help researchers detect animals in images and identify specific animals within a species. Researchers can monitor animals such as birds, amphibians, cetaceans, and even fish and analyze the data through machine-learning tools.
With these technologies, scientists can make:
- Better habitat analysis.
- More precise guesses on wildlife and species.
- Analyze climate change’s impact on animals in real-time.
Real-life example: Wildbook uses neural networks and computer vision algorithms to find and count animals in pictures and to identify individual animals within a group. With this knowledge, wildlife population sizes can be estimated more accurately.
Wildbook’s vision on sustainability.
Data analysis for sustainability
Large language models (LLMs) like GPTs are crucial in driving a more sustainable future by assisting organizations in analyzing and taking action based on large datasets. Some key applications of AI in this domain include:
9. Analyzing business documents & reducing waste
Generative AI systems can review and analyze business documents, helping companies spot opportunities to cut waste and improve their sustainability efforts. For example:
- Generative AI tools can analyze data on transportation, energy use, and other resource consumption to provide accurate carbon footprint calculations at a lower cost.
- AI algorithms can optimize supply chain processes by identifying inefficiencies and suggesting ways to cut fuel consumption. These technologies help reduce greenhouse gas emissions and minimize resource use.
- By leveraging AI, companies can gain valuable insights into their energy consumption, helping them shift to renewable energy sources and improve overall energy efficiency.
This integration of AI technologies allows businesses to reduce their environmental impact while embedding sustainability into their operations.
10. Identifying scope three risks
Detecting Scope 3 greenhouse gas emissions, those generated indirectly through supply chains and product lifecycles, can be challenging. However, by using AI tools like ChatGPT, companies can effectively identify these risks by analyzing large volumes of publicly available data, such as:
- News articles, industry reports, and social media posts that highlight environmental challenges related to suppliers or production processes.
- Emerging environmental sustainability risks that might impact sustainability strategies.
Businesses can proactively address climate change concerns and align with environmental justice principles by identifying these risks.
11. AI for energy and resource optimization
AI systems, including those deployed by cloud service providers, can help businesses and organizations:
- Optimize energy usage in data centers through better cooling systems, improving power usage effectiveness (PUE).
- Predict and manage energy storage needs, aligning renewable energy generation with demand.
- Reduce electronic waste by extending the entire lifecycle of devices using AI-driven maintenance recommendations.
Real-life example: NVIDIA’s Earth‑2 is a GPU‑accelerated climate‑simulation platform enabling kilometre‑scale global modelling.
It launched a generative‑AI model called cBottle (“Climate in a Bottle”) in June 2025. The model can generate global atmospheric states conditioned on inputs like time of day and sea‑surface temperatures, with resolution down to 1‑2 km and significantly reduced computation time and energy use.
This system achieves:
- Data compression ratios up to 3,000× per sample.
- Forecast speeds thousands of times faster and up to 10,000× more energy‑efficient than traditional methods.
- Integration of AI‑based downscaling (CorrDiff) to provide super‑resolution weather insights.
- Active adoption by leading research institutions (MPI‑M, AI², Alan Turing Institute) facilitates interactive digital‑twin climate exploration.
Key features include:
- Kilometre‑scale climate simulation and interactive visualisation.
- Generative AI (cBottle + CorrDiff) for rapid, high‑resolution forecasts.
- Proven by real‑world testing (GTC, hackathons) and institutional collaboration.
Sustainable agriculture
AI technologies in agriculture are helping farmers manage challenges like resource inefficiency and environmental impact. By incorporating tools such as agricultural robotics, weather monitoring systems, and land management algorithms, farmers can optimize their operations, reduce waste, and meet sustainability goals.
Additionally, AI-powered crop and animal monitoring ensures healthier yields and livestock by detecting issues early, reducing the need for chemicals, and minimizing resource usage.
12. Agricultural robotics
Similar to a self-driving car, AI-powered robots can move around and harvest crops when they’re ready and mature enough. This helps reduce waste and can improve production lines.
13. Weather monitoring
AI can also monitor and forecast the weather. This helps farmers predict the weather in a specific location, giving them insights into when to water their crops or when it’s the best time to plant or harvest.
14. Land management
Another use case of AI is farmland planning. Using satellite images, algorithms, and land-use data, farmers can plan where and when to plant their crops. This can also help them ensure regulatory compliance.
15. Crop and animal monitoring
AI can help farmers keep their crops and animals healthy. With image recognition and sensors to spot crop conditions, AI can help reduce bugs attacking crops or early signs of animal diseases.
Farmers can then step in and fix the problem without using excessive amounts of chemicals or medicines, reducing potential losses.
Sustainable production and workplace
16. Less defective production
AI-enabled computer vision systems can address product return issues stemming from defects or customer dissatisfaction by minimizing production errors at the manufacturing stage.
Computer vision-enabled quality control systems installed on the conveyor belt or production line can inspect the quality of the product more accurately and efficiently than manual inspection.
See how it works:
Video on AI Visual Inspection to detect defects in manufacturing lines.
This reduction in defective products can ultimately reduce the organization’s product returns and GHG emissions related to reverse logistics and other return processes.
17. Better leak detection in production
Computer vision systems can help detect water leaks and other harmful chemicals within a production plant and alert authorities to take quick action. This can help businesses reduce their environmental impact.
See how it works:
Video on leak detection systems to achieve safer, more accurate, and cost-effective remote site monitoring.
18. Safer workplace
Sustainability consists of three parts: environmental, social, and governmental. To be truly sustainable, a business needs to focus on all three.
AI-enabled computer vision systems can help improve worker safety by ensuring compliance with safety rules. This can help improve a business’s social sustainability by making the business more secure for its workers.
Smart cameras can be installed at certain points of the manufacturing facility to inspect if the workers are following the rules and wearing safety equipment. The system can also identify other risks in the facility and notify the relevant operations or safety manager for further action.
Energy and logistics
19. Reduced energy consumption
Figure 2: Global share of electricity from renewable resources.
Even though investments in renewable energy have increased significantly in the past few years, renewable energy accounts for only 30% of the world’s electricity generation.
AI can help increase the use of renewable energy by studying the patterns of energy consumption and providing insights on reducing and improving consumption while not compromising the company’s productivity.
20. Optimized and sustainable logistics
AI can also help improve the sustainability of the distribution and logistics operations of a business, which account for a significant chunk of the total corporate carbon footprint.
AI-powered software can optimize product-delivery routes by incorporating sustainability as a key factor. Route optimization systems have become a necessity for logistics firms, as they offer significant financial and environmental benefits.
Watch how AI and digital twin technologies are helping with sustainable last-mile delivery:
AI and digital twin technologies to help with sustainable last-mile delivery.
Check out logistics AI use cases to learn more about how AI is revolutionizing the logistics sector.
What are the challenges of sustainability AI?
Artificial intelligence looks promising in helping protect the environment, but it also presents some challenges:
Computing energy
Advanced AI models need significant computing power, which means they use a lot of energy.
This influences both operational prices and carbon emissions. Thus, using energy-intensive AI technologies in the service of environmental sustainability can be paradoxical.
Labor abuses
Large language models such as ChatGPT can require labels to keep the model away from toxic texts. To get these labels, OpenAI sent tens of thousands of particle texts to a firm in Kenya. The data labelers employed by the company are paid only around $1.32 and $2 per hour.
This raises questions about whether workers’ rights are being usurped in developing AI tools for a sustainable future.
AI bias and ethics
AI models learn from data, and if the data is biased or represents only a particular part of reality, the models can produce incorrect results. For instance, an AI model trained in location-specific data can fail to generate data for other areas.
Decisions based on AI results can greatly affect society and the world. Therefore, questions can arise about data privacy and ownership.
Best practices to mitigate challenges
Energy-efficient AI
The priority should be using algorithms and devices that use less energy. Research groups can work on designing models that balance how well AI works and how much energy it uses.
AI computing infrastructure can be powered by renewable energy sources, which can help lower the carbon footprint even more.
Addressing AI bias
AI models should use appropriate methods for collecting, testing, and validating data to avoid bias. Including representative data and considering how conditions can vary in different locations are also important.
Developing ethical guidelines
For AI to protect the environment, ethical guidelines and policies must be designed and followed. This includes clear rules about who owns the data, how to keep it private, and how to use AI ethically.
Encouraging stakeholder engagement
Involve stakeholders in the decision-making process, specifically the groups that will be affected by AI’s results. This means ensuring everyone knows how AI models work and what data they use.
Principal Analyst
Cem Dilmegani
Principal Analyst
Cem’s work has been cited by leading global publications including Business Insider, Forbes, Washington Post, global firms like Deloitte, HPE and NGOs like World Economic Forum and supranational organizations like European Commission. You can see more reputable companies and resources that referenced AIMultiple.
Throughout his career, Cem served as a tech consultant, tech buyer and tech entrepreneur. He advised enterprises on their technology decisions at McKinsey & Company and Altman Solon for more than a decade. He also published a McKinsey report on digitalization.
He led technology strategy and procurement of a telco while reporting to the CEO. He has also led commercial growth of deep tech company Hypatos that reached a 7 digit annual recurring revenue and a 9 digit valuation from 0 within 2 years. Cem’s work in Hypatos was covered by leading technology publications like TechCrunch and Business Insider.
Cem regularly speaks at international technology conferences. He graduated from Bogazici University as a computer engineer and holds an MBA from Columbia Business School.
View Full Profile
Researched by
Sıla Ermut
Industry Analyst
Sıla Ermut is an industry analyst at AIMultiple focused on email marketing and sales videos. She previously worked as a recruiter in project management and consulting firms. Sıla holds a Master of Science degree in Social Psychology and a Bachelor of Arts degree in International Relations.
View Full Profile
