Build an AI Sustainability Strategy

Learning Objectives

After completing this unit, you’ll be able to:

Articulate the value of creating an AI sustainability strategy.
Explain the preliminary approach to AI sustainability developed by Salesforce.
Implement strategies for AI sustainability at your organization.

Why Do You Need an AI Sustainability Strategy?

While AI shows promise as a powerful tool to accelerate sustainability solutions, it also presents potential environmental risks.

The large language models (LLMs) that power generative AI require significant compute resources to function, which can result in negative environmental impacts, including energy, water, and natural resource demands. We believe every business developing and deploying AI needs a strategy to implement ethical and sustainable technology from the start. Let’s explore how to do this.

AI Sustainability: A 3-Pillar Framework

Sustainability is a guiding principle for the development and deployment of AI here at Salesforce. Our preliminary approach, developed in close partnership across Salesforce AI Research, Sustainability, and the Office of Ethical and Humane Use, focuses on three core pillars: smart demand, efficiency, and clean supply.

The Salesforce three-pillar AI sustainability framework focuses on smart demand, efficiency, and clean supply.

Let’s dive in to what each of these mean.

Smart Demand

This means using AI wisely. Think of AI as a precision instrument—not every AI tool is the best fit for every task. Begin each project by asking how much AI, and of what type, is needed for success.

Here’s what smart demand looks like in action:

Design products wisely: Avoid unnecessary AI usage, use simpler techniques or algorithms, default to efficient models, and consider architectural approaches that solve challenges with less AI.
Use right-sized models: For many tasks, a compact, specialized model may offer similar accuracy to a frontier-scale system, at a fraction of the energy cost.
Be transparent and flexible: Give users visibility into AI’s resource impact and foster flexibility around when and how data centers consume power.
Align incentives: Align cost structures to encourage efficiency. Usage-based pricing aligns pricing incentives with sustainability. Because costs map directly to compute, customers enjoy immediate financial upside when they reduce emissions.

For example…

At Salesforce, managing demand means partnering closely with our customers—educating, empowering, and equipping them with the right tools at the right time to make smart, sustainable choices. Agentforce is built to deliver high performance while aiming to mitigate environmental impact, unlike DIY AI approaches that require energy-intensive model training for each customer.

Our agentic architecture moves beyond a single, general LLM, enabling the use of smaller, more efficient AI models where possible. And we have a flexible, usage-based pricing structure for Agentforce—so when customers reduce their usage, they lower both their costs and their associated environmental footprint. We promote transparency in energy consumption through initiatives like the AI Energy Score, and offer educational resources via Trailhead to foster sustainable practices in our ecosystem and beyond.

Efficiency

This means doing more with less. Once you know what intelligence is needed, efficiency focuses on how to deliver it. Efficiency techniques can compound: Small savings at the model, code, data, and hardware layers can add up to big reductions in energy, carbon, water, and cost.

Here’s what efficiency looks like in action:

Develop and deploy efficient AI models: Smaller, purpose-built models trained for specific domains can outperform large general-purpose models in both speed and energy efficiency.
Use model optimization techniques: Quantization, distillation, and pruning techniques can help streamline models for lower compute without sacrificing quality.
Optimize inputs: Concise, well-structured prompts can use less energy and deliver faster results.
Adopt green software practices: Think beyond the model. Optimize code for energy use, reduce idle compute time, and avoid unnecessary layers of abstraction.
Consider total data center efficiency: Beyond individual components, it's essential to assess the efficiency of the data centers you use.
Shift AI to the edge: By running smaller, more efficient AI models on smaller devices like laptops and phones, some computational load can be moved away from large data centers—saving energy and improving efficiency.

For example…

At Salesforce, our AI Research team is pioneering domain-specific, efficient AI models, designed to excel at specific tasks while consuming less energy than large-scale LLMs. We test new paradigms of AI model usage, enabling small models to run locally, which can avoid energy-intensive cloud data centers. And Agentforce uses deeply integrated, structured data from across the Salesforce Platform, ensuring high accuracy while minimizing compute.

We also “think beyond the model” with our Green Code initiative and Sustainability Guide for Salesforce Technology. And we evaluate data centers using key metrics, including Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE), to help ensure we partner with data center suppliers that operate efficiently.

Clean Supply

This means powering AI with low-impact resources. No matter how efficient your systems are, they still need resources: energy, water, and other natural resources. Clean supply focuses on making sure that these inputs come from more sustainable sources.

Here’s what clean supply looks like in action:

Sourcing from more sustainable suppliers: AI supply chains are vast, involving cloud providers, chip manufacturers, infrastructure operators, and software vendors. Embedding sustainability across procurement can drive change far beyond your own footprint.
Invest in clean power and water: AI sustainability depends on scaling clean power and resilient watersheds to serve it.
Advocate for systemic change: Engage with regulators and industry groups to shape the policies and infrastructure that will govern AI’s environmental impact.

For Example…

At Salesforce, we invest in the energy transition and embed sustainability across our operations. This includes active investments in a clean energy future through initiatives like our Ventures Impact Fund, philanthropic grants, and Salesforce Accelerator.

We ask partners across our supply chain to decarbonize and provide resources like our Net Zero Toolkit to aid their journey. And we advocate for systemic change by collaborating with communities, regulators, and industry groups to advance sustainable AI practices, including through our Sustainable AI Policy Priorities and by joining the Coalition for Sustainable AI.

A Holistic Approach to AI Sustainability

Sustainability is one of our core values at Salesforce. This means it guides actions we take across the business, including how we develop and use technology like AI.

In addition to prioritizing sustainable AI development, we also use AI to accelerate sustainability solutions. Through the Salesforce Accelerator—Agents for Impact, we provide nonprofits with funding, technology, and pro bono expertise to help them scale solutions to address climate challenges with Agentforce.

With Agentforce for Net Zero Cloud, we help sustainability teams streamline reporting, reduce operational costs, and boost efficiency with pre-built agent topics and actions. Grounded in company data and built on the Salesforce Trust Layer, Agentforce augments sustainability teams with digital labor—a workforce of intelligent AI agents that work alongside humans to complete time-intensive tasks faster and more efficiently.

There are so many exciting possibilities for sustainable development and using AI for a more sustainable future. By sharing our learnings on AI sustainability so far, we hope you join us on the journey.

Geschätzte Zeit

Themen

Benötigen Sie Hilfe?

Einstein – Ressourcen