AI holds unimaginable potential to remodel our world, drive innovation, and be a strong software for sustainability, nevertheless it additionally poses potential environmental challenges, together with elevated vitality, water, and pure assets demand.
Analysts warn that by 2027, 40% of AI data-center initiatives may hit energy bottlenecks, and international AI demand may require as a lot as 4-6x the annual water withdrawal of Denmark. These potential strains threaten delays in technological innovation, greater prices, disruptions in energy reliability, and elevated competitors for recent water.
In our current AI Sustainability Outlook, we element Salesforce’s ongoing work to grasp and scale back the environmental affect of AI and ship on its potential for good. Whereas we’re dedicated to main the best way, we additionally know we are able to’t do that alone. We imagine each enterprise must implement sustainability into their AI technique from the beginning.
Beneath, you’ll discover our preliminary 3-pillar framework for AI sustainability, specializing in sensible demand, effectivity, and clear provide. Then we’ll clarify how Salesforce prospects can apply it to their very own utilization.
Let’s dive in:
Salesforce AI Sustainability Outlook
Discover our ongoing work to grasp and scale back the environmental affect of AI and ship on its potential for good.
Developed in shut partnership throughout Salesforce AI Analysis, Sustainability, and the Workplace of Moral and Humane Use, our 3-pillar framework intends to information the AI ecosystem towards sustainability. The framework is deliberately versatile, so anybody can put it to work, and continues to evolve to accommodate the speedy tempo of AI innovation.
Salesforce’s preliminary 3-pillar framework for AI sustainability focuses on sensible demand, effectivity, and clear provide.
1. Good demand: Use AI properly
AI sustainability begins with sensible demand. Begin each mission by asking how a lot AI, and what sort of AI, is required for fulfillment. Consider AI as a precision instrument — with completely different instruments used for particular duties.
As you be taught which AI instruments are best and related to your work, share your data together with your colleagues and prospects. Educate, empower, and equip them with the correct instruments on the proper time to make sensible, sustainable decisions.
For instance:
Use AI mindfully by way of monitoring and optimization: Use the Agentforce Command Middle and Digital Pockets to trace utilization and guarantee optimum efficiency.
Use right-sized fashions: The Agentforce structure is designed to use the correct degree of intelligence on the proper time. For instance, inside the Agentforce Belief Layer, we’ve finetuned environment friendly fashions for Toxicity Detection, Personally Identifiable Data (PII) Masking, and Immediate Protection. These fashions vary from 44 to 135 million parameters in measurement — about 99% extra environment friendly than frontier LLMs.
2. Effectivity: Do extra with much less
As soon as what AI instruments are wanted, effectivity focuses on the best way to ship it. Effectivity strategies can compound: small financial savings on the mannequin, code, information, and {hardware} layers may add as much as large reductions in vitality, carbon, water, and value. You don’t should deal with each approach without delay. Begin small. Break it down, take heed to your stakeholders, and keep constant. Transparency round learnings as you go will earn the belief of your stakeholders.
For instance:
Choose environment friendly AI fashions: Selecting extra environment friendly AI fashions could make an enormous distinction in your footprint. With Salesforce, you’ll be able to select completely different fashions by way of Immediate Builder and Agentforce. We now have additionally included sustainability metrics in our Agentic Benchmark for CRM to allow you to grasp the environmental tradeoffs between fashions.
Optimize inputs: Concise and well-structured prompts use much less vitality and ship quicker outcomes. Curated, task-relevant information can scale back the necessity for large-scale inputs and enhance mannequin precision. Take a look at the Immediate Design Information for finest practices and ideas.
Undertake inexperienced software program practices: Optimize code for vitality use, scale back idle compute time, and keep away from pointless layers of abstraction. See Salesforce’s Inexperienced Code initiative and Sustainability Information for Salesforce Know-how for extra.
3. Clear provide: Energy AI with low-impact assets
Regardless of how environment friendly your methods are, they nonetheless require assets like vitality, water, and uncooked supplies to operate. Clear provide focuses on ensuring these assets come from extra sustainable sources.
This pillar extends past inner operations, addressing broader system-level shifts throughout grids, provide chains, and public coverage.
For instance:
Select greener information heart areas: Hyperforce — Salesforce’s public cloud information heart infrastructure — lets you deploy in areas internationally. Completely different areas can have large variations by way of their carbon depth and water footprint, and shifting to a greener area could make an enormous distinction.
Demand transparency from AI suppliers: Transparency is foundational to allow customers to decide on extra sustainable fashions. The AI Power Rating, co-developed by Salesforce, allows AI builders to reveal this information. Salesforce is dedicated to disclosing this information, and you may make an affect by asking your AI suppliers for this information, too.
Supply from extra sustainable suppliers: AI provide chains are huge, involving cloud suppliers, chip producers, infrastructure operators, and software program distributors. Embedding sustainability throughout procurement can drive change far past your individual footprint. Salesforce’s Sustainability Exhibit is a good instance of how one can ask suppliers to set science-based targets, enhance sustainability disclosures, and ship carbon-neutral services and products — all as a part of their contractual obligations.
Spend money on clear energy and water: Electrical energy is the lifeblood of AI. Clear vitality applied sciences, like photo voltaic, batteries, and geothermal can assist safe inexpensive, considerable, and dependable energy for everybody, whereas additionally lowering impacts on human and environmental well being. Speed up the worldwide transition to scrub vitality and improve watershed resilience by procuring high-quality renewable vitality, supporting nature-based options that shield watersheds and ecosystems, and actively investing in a clear vitality future by way of philanthropic grants.
Advocate for systemic change: Interact with regulators and business teams to form insurance policies that govern AI’s environmental affect — for instance, supporting clear vitality infrastructure, emissions transparency, and accountability mechanisms.
AI remains to be in its infancy, and our understanding of its potential dangers and advantages will proceed to develop. However even at this stage, it’s clear that these challenges require us to take motion.
We imagine an AI sustainability technique that mixes sensible demand, effectivity, and clear provide can assist make sure that AI grows in a accountable and sustainable method. Specializing in only one pillar in isolation — like effectivity — can result in rebound results and missed alternatives for systemic change. It’s vital to take a holistic method.
At Salesforce, we’re pleased with the progress we’ve made, however we all know there’s extra work forward. Collective motion has by no means been extra essential, and we hope you’ll be a part of us.
Sustainability, simplified with Agentforce.
Improve effectivity whereas lowering environmental affect with Agentforce for Web Zero Cloud.
