Artificial intelligence (AI) has long been the subject of suspicion, mistrust, and even fear, but lately its reputation is suffering from reports about the negative impact it can have on the environment, from the amount of energy and water it uses to the harmful emissions it creates.

Our teams at Slalom, however, believe there are ways to mitigate these negative effects. We recently published a discussion on the oddly contradictory idea that generative AI (GenAI) may hold the key to its own redemption (in terms of environmental impact, at least). But we felt it would also help to dig deeper into what the tech companies themselves are doing, both to make AI more sustainable and to help their customers operate more sustainably. Microsoft—a market leader in GenAI—provides a great example of how to address this complex problem.

Generative AI requires tremendous amounts of energy, both for the tech companies to develop and host the foundation models that form the basis of GenAI computations and for users who run GenAI workloads. Goldman Sachs reports, for example, that a single ChatGPT query requires roughly 10 times the energy of a Google search. The data centers that process these workloads also consume vast amounts of water—an estimated 4.2 billion to 6.6 billion cubic meters by 2027—to run cooling systems. Harmful emissions are the other side of the coin. Microsoft reported that its carbon emissions have increased by 30% since 2020, largely from the construction of data centers to support generative AI.

The recognition that this trajectory is unsustainable has led Microsoft to adapt in a couple of different ways:

• Using cleaner, more renewable energy and water sources: Growing the availability of renewable energy, both for itself and for the regions in which it operates, is a priority according to Melanie Nakagawa, Microsoft's chief sustainability officer. The company is prioritizing a mix of net new, clean power sources and renewable energy credits while also exploring investments in nuclear.
  
  It’s also working on ways to put energy back into local grids from its data centers, for example by capturing surplus wind energy production in Ireland and directing excess heat from data centers to local homes in Denmark. Harvesting rainwater, reclaiming water from utilities, and using adaptive ventilation systems to cool data centers are a few ways Microsoft is minimizing its reliance on freshwater consumption.
  
  
• Making the technology itself more sustainable: It’s not enough to simply use cleaner energy. The amount of energy and natural resources consumed must also be reduced. This is why Microsoft has been launching small language models, or SLMs, such as its Phi series, which have met or exceeded the performance of much larger and more energy-intensive models.
  
  In partnership with the Department of Energy’s Pacific Northwest National Laboratory (PNNL), Microsoft is also helping to identify new, more-sustainable battery material. Using Azure Quantum Elements—a product designed to accelerate scientific discovery—Microsoft and PNNL screened over 32 million materials and identified over 500,000 stable candidates with a process that took weeks, rather than years.

What’s clear from the example above is that GenAI can be a formidable ally in the effort to operate more sustainably. At a digital event held in early 2024, aptly called This Is AI … for Sustainability, Microsoft executives explained how AI can help organizations achieve their energy and emissions goals by making it easier for them to:

• Measure, predict, and optimize systems that are far too complex for traditional computational systems
• Accelerate the discovery and development of sustainability solutions, such as low-carbon materials, renewable energy production and storage, and climate-resilient crops
• Empower the sustainability workforce by digesting huge volumes of data to help leaders make better-informed decisions

AI’s ability to process vast amounts of information from multiple sources is a game changer for a data-intensive field like sustainability. During the digital event, Satish Thomas, corporate VP for Microsoft Industry Clouds, explained that organizations must first harmonize their data in a centralized data estate. By doing this, all sustainability data — for scope 1, 2, and 3 emissions — can be combined with other enterprise data and analyzed together to achieve more insightful and actionable analyses. With Microsoft Cloud for Sustainability (including Microsoft Sustainability Manager) and Microsoft Fabric working together, you have a powerful solution for all data analytics needs. This pairing creates a single place for you to gather, store, analyze, and visualize sustainability data, regardless of the source.

Once your data is in order, it’s time to unleash the power of GenAI to get at those vital insights. Using Copilot in Microsoft Sustainability Manager, you can upload additional sustainability-related documents, such as the Greenhouse Gas Protocol’s Scope 3 Calculation Guidance, combine them with your internal sustainability data, and ask questions of the data in natural language—so you don’t need to be an experienced data scientist or expert in mathematical modeling. The copilot can also create reports, build calculation models, and summarize or answer questions on large amounts of unstructured data. This is a much faster and easier way to identify outliers, trends, and correlations in the data. You can also explore what-if scenarios to go beyond historical data analysis to more predictive scenario modeling.

Copilot in Microsoft Sustainability Manager, notes Mandar Zope, Slalom senior principle, Microsoft Business Applications, “can help save a lot of time and resources, because now you can actually ask the data things like ‘What am I missing?’ or ‘What am I doing wrong?’ rather than having humans conduct the tedious analysis.” What’s more, Zope adds, is that “Microsoft also provides all the different Azure resources your company might be using embedded in that tool,” so you can work with the most comprehensive data set possible when conducting analyses and formulating an action plan.

As more and more regional, national, and global regulations governing energy, emissions, and sustainable practices are enacted, organizations will have to respond by updating processes, improving — or perhaps instituting — reporting capabilities, or even completely overhauling their business models to fall in line with new guidelines. With the leading perspective on Microsoft Fabric and generative AI, Slalom can help you create a unified analytics solution across every aspect of the Microsoft data platform — including sustainability data.