Does your AI Strategy Include Sustainability? Here’s Why It Should

Sunya Norman

Salesforce AI Sustainability Outlook

Explore our ongoing work to understand and reduce the environmental impact of AI and deliver on its potential for good.

1. Smart demand: Use AI wisely

AI sustainability begins with smart demand. Start every project by asking how much AI, and what type of AI, is needed for success. Think of AI as a precision instrument - with different tools used for specific tasks.

As you learn which AI tools are most effective and relevant to your work, share your knowledge with your colleagues and customers. Educate, empower, and equip them with the right tools at the right time to make smart, sustainable choices.

For example:

Use AI mindfully through monitoring and optimization: Use the Agentforce Command Center and Digital Wallet to track usage and ensure optimal performance.

Use right-sized models: The Agentforce architecture is designed to apply the right level of intelligence at the right time. For example, within the Agentforce Trust Layer, we've finetuned efficient models for Toxicity Detection, Personally Identifiable Information (PII) Masking, and Prompt Defense. These models range from 44 to 135 million parameters in size - about 99% more efficient than frontier LLMs.

2. Efficiency: Do more with less

Once you know what AI tools are needed, efficiency focuses on how to deliver it. Efficiency techniques can compound: small savings at the model, code, data, and hardware layers could add up to big reductions in energy, carbon, water, and cost. You don't have to tackle every technique at once. Start small. Break it down, listen to your stakeholders, and stay consistent. Transparency around learnings as you go will earn the trust of your stakeholders.

For example:

Select efficient AI models: Choosing more efficient AI models can make a big difference in your footprint. With Salesforce, you can choose different models through Prompt Builder and Agentforce. We have also included sustainability metrics in our Agentic Benchmark for CRM to enable you to understand the environmental tradeoffs between models.

Optimize inputs: Concise and well-structured prompts use less energy and deliver faster results. Curated, task-relevant data can reduce the need for large-scale inputs and improve model precision. Check out the Prompt Design Guide for best practices and tips.

Adopt green software practices: Optimize code for energy use, reduce idle compute time, and avoid unnecessary layers of abstraction. See Salesforce's Green Code initiative and Sustainability Guide for Salesforce Technology for more.

3. Clean supply: Power AI with low-impact resources

No matter how efficient your systems are, they still require resources like energy, water, and raw materials to function. Clean supply focuses on making sure these resources come from more sustainable sources.

This pillar extends beyond internal operations, addressing broader system-level shifts across grids, supply chains, and public policy.

For example:

Choose greener data center regions: Hyperforce - Salesforce's public cloud data center infrastructure - enables you to deploy in regions across the world. Different regions can have big differences in terms of their carbon intensity and water footprint, and shifting to a greener region can make a big difference.

Demand transparency from AI suppliers: Transparency is foundational to enable users to choose more sustainable models. The AI Energy Score, co-developed by Salesforce, enables AI developers to disclose this data. Salesforce is committed to disclosing this data, and you can make an impact by asking your AI suppliers for this data, too.

Source 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. Salesforce's Sustainability Exhibit is a great example of how you can ask suppliers to set science-based targets, increase sustainability disclosures, and deliver carbon-neutral products and services - all as part of their contractual obligations.

Invest in clean power and water: Electricity is the lifeblood of AI. Clean energy technologies, like solar, batteries, and geothermal can help secure affordable, abundant, and reliable power for everyone, while also reducing impacts on human and environmental health. Accelerate the global transition to clean energy and enhance watershed resilience by procuring high-quality renewable energy, supporting nature-based solutions that protect watersheds and ecosystems, and actively investing in a clean energy future through philanthropic grants.

Advocate for systemic change: Engage with regulators and industry groups to shape policies that govern AI's environmental impact - for example, supporting clean energy infrastructure, emissions transparency, and accountability mechanisms.