Artificial intelligence and sustainability are hot topics in business, but while AI has enormous and accelerating momentum, there is concern that sustainability’s moment may be passing. In truth, both are profoundly important and in only their very first stages. Indeed, we are early enough in their evolution to bring AI and sustainability together to create—using what we call an eco-AI approach—an incredibly powerful source of advancement for both the planet and the corporate bottom line.

Four pioneering strategic applications of AI

Increasing numbers of forward-thinking companies are using AI to work on sustainability in ways that generate true business value. Here are four approaches that are worth every CEO and business leader’s attention.

Deliver value to customers while boosting sustainability. Consumers and customers continue to rate sustainability as an important purchase criterion, but they often lack a clear understanding of what makes a product or service sustainable. AI can help close this gap by providing new and more effective approaches to communicate about sustainable products and propositions. Home furnishings giant Ikea, for example, built an AI recommendation engine that can tailor product suggestions for consumers based on their sustainability preferences. Twenty percent of interactions with the tool drive traffic to the company’s website, with 5% of those visits leading to transactions.   

Improve financial and sustainability results. AI and digital systems can help companies develop sustainable offerings that save money, streamline innovation, and build new businesses. Consider how a food company might use digital tools to track and reward farmers for reducing their emissions. In addition to creating a more sustainable supply of raw materials, this could help the company build a premium, low-carbon product line. Profits from that line could then be used to pay back the upfront costs, thereby creating a positive flywheel.

Reduce operational risk and maximize resilience. Bain & Company estimates that losses from natural disasters could represent up to 4% of global GDP by 2050. Remote monitoring, space-based technology, and more powerful predictive models all will be needed to assess exposure and build more resilient operations. AI can help mining, agriculture, and other companies estimate the exposure of facilities to a range of natural risks, including precipitation, heat, fire, wind, cold, and flood, and develop mitigation and transition plans for the most endangered locations based on that analysis.

Build operational and supply chain digital twins. When fully deployed, AI will revolutionize how companies identify and realize sustainability improvements within operations and along the supply chain. By instantly modeling the impact of decisions on spending, carbon emissions, and other sustainability metrics, digital twins will strengthen decision making and reduce consumption of materials, energy, and water. In the public sector, the Virtual Singapore platform shows the possibilities. Pulling from diverse data sources, the platform’s 3D city model helps urban planners and designers identify opportunities for energy efficiency, assess the environmental impacts of development, and reduce emissions through optimized transportation systems.

Eco-AI’s power couple: CSOs and CTOs

As AI experimentation accelerates, leaders must consider the future implications of their IT strategies and priorities on their net-zero plans. While AI holds great promise to advance and speed sustainability efforts, the potential impact on emissions must be understood and addressed from the outset. This will require companies’ sustainability and technology functions to work together on key priorities. Three principles for effectively doing so are emerging.

1. Technology’s power use and emissions can no longer be an afterthought. Carbon emissions from IT traditionally have been seen as little more than a rounding error. This was fine when most companies’ IT departments had a relatively small carbon footprint. For a typical consumer products company, for example, IT has historically represented about 1% of its carbon footprint, compared with 25% each for packaging and raw materials.

AI will change that. By 2030, Bain projects that the growth of AI, along with increased cloud usage and rising volumes of data in traditional applications, will lead to significantly higher IT carbon emissions across industries. In consumer products alone, IT emissions are expected to increase by at least three times (see Figure 1).

Figure 1

The carbon emissions of IT departments are climbing, driven by generative AI

A number of factors are pushing up AI’s energy use and carbon emissions. The first is an explosion of users and applications. Bain’s 2024 cross-industry AI survey shows that almost 90% of large companies in the US are using generative AI in some capacity. And corporate users are tapping increasingly large, sophisticated, and power-hungry models. The total number of parameters grew from 1.5 billion for GPT-2 in 2019 to 1.7 trillion for GPT-4 in 2023. Users are also engaged in more energy-intensive activities. Video generation, for example, requires up to 300 times more power than image generation. Projected gains in the energy efficiency of processes and chips are unlikely to offset this surging demand.

A significant portion of tech-related emissions sits outside the IT department’s control. As a result, companies will have to take a broad and systemic view when mapping future emissions from technology. For B2C companies, this should include a calculation of emissions from consumer use of AI-enabled apps. With a large user base and a bias toward image and video generation, many of these apps are more power-intensive than most organizations realize. As AI-enabled initiatives expand in teams like purchasing, marketing, and finance, CTOs and CSOs will have to work across functions to get the information needed to build a holistic picture. This review should include work with third-party suppliers—for example, when marketing departments work with outside agencies on AI use cases.

2. Win the race to decarbonize your cloud. Bain analysis shows that up to 70% of a typical company’s IT Scope 3 decarbonization goals will depend on the decarbonization of its IT suppliers. The fastest way to decarbonize IT is therefore to engage suppliers and support their decarbonization journeys. Cloud providers are scrambling to meet the burgeoning demand for sustainably powered data services and to capture the opportunities this presents. While progress is advancing on many fronts, the supply of green energy will be quite constrained in the medium term, creating significant headwinds for the net-zero ambitions of both the tech sector and its customers. Indeed, many experts expect total demand for electricity to increase beyond total supply in the next few years (see Figure 2).

Figure 2

If electricity generation growth remains as forecast, demand is likely to exceed supply in the near term

Companies should act now to ensure they are first in line for green power. This starts with understanding the power usage efficiency of data center providers and alternative suppliers. There is a high level of variation across companies and among the sites of any single provider. New tools to track the energy efficiency of cloud service providers are becoming available, and suppliers are offering dashboards to help monitor and test consumption. Supplier selection and management processes will need to be bolstered, and purchasing teams will have to be trained to embed sustainability criteria in their process.

3. Don’t hit the brakes on AI, but integrate sustainable behavior from the start. Could the combination of tremendous growth in AI demand, limited availability of green energy, and sustained stakeholder pressure to decarbonize eventually result in constraints on AI usage, or even rationing? While this sounds extreme, it’s urgent that forward-looking CEOs, CTOs, and CSOs push their organizations to use AI in the most effective and efficient ways.

Two areas warrant immediate focus. The first is upskilling and creating awareness within the organization of eco-design and eco-utilization of generative AI. This includes selecting appropriately sized models for the task at hand. There can be more than 100 times difference in power use between the smallest and the largest model when applied to the same task. Also, not everyone needs AI’s most powerful tools. Bain estimates, for example, that 90% or more of a typical consumer goods company’s employees do not require access to energy-consuming video generation tools.

There are other technical actions to take as well. Companies can leverage prompt engineering, such as the selection of predefined prompts for all users and semantic “caching” of existing responses based on identical or similar user requests, to reduce the number of requests per user. Fine-tuning a model rather than using a multipurpose one can reduce emissions by up to 70%, and deploying quantization in open-weight models to reduce model size and speed up processing can reduce emissions by up to 50% without a significant impact on output quality. 

* * *

Merging AI and sustainability presents tremendous opportunities for business. Companies should strategically embed AI within sustainability initiatives to fuel innovation, efficiency, and resilience. However, the surge in AI’s energy demand calls for a smart, sustainable approach. By embedding sustainability from the start, businesses can meet carbon targets and lead the charge toward a greener, tech-driven future. The challenge is clear: Innovate fast, but do it sustainably.