Two major corporate priorities—artificial intelligence and sustainability—have begun to intersect.
Though AI may have tremendous implications for environmental impact programs in the future, some executives remain skeptical.
New research from Salesforce shows that while six in 10 sustainability professionals believe AI’s benefits are likely to outshine its risks when used as a tool for climate crisis progress, four in 10 have fears that the technology will negatively impact their organizations’ sustainability efforts.
That could be due to the fact that training foundation models, like OpenAI’s GPT-4, for instance, requires a great deal of energy. The applications built on top of those foundation models, too, could be increasing carbon emissions, especially in industries like retail and apparel, where consumer-facing systems may be receiving thousands of queries a day, as compared to an internal system that handles just a few hundred queries a week.
In fact, 81 percent of those surveyed indicated that decreasing AI’s carbon footprint is a priority, with one in three sustainability leaders noting that goal remains a high priority inside their organizations.
With data from the International Energy Agency showing that the power needed to keep up with the growth of data centers could double by 2026, sustainability professionals’ concern over the energy footprint makes sense.
But the fact that the impacts of AI systems will also fall into companies’ Scope 3 emissions reporting could also be playing a role, said Boris Gamazaychikov, senior manager of emissions reduction for Salesforce.
“Unless your company is directly training these AI models, the emissions associated are likely going to go into Scope 3…and with regulations like CSRD, it’s going to be more and more important to have a great grasp on your Scope 3 emissions and have reduction strategies over time,” he explained.
As many companies struggle to control their Scope 3 emissions because of supplier contracts and supply chain issues, adding another energy-hungry tool to the mix will likely be a strategic balance.
Despite some of the worries leaders have, about half of surveyed leaders are still using the technology to work toward climate-related goals in one way or another.
One in five surveyed professionals said they have fully implemented AI into the heart of their sustainability programs and initiatives. That seems to be on track with other business functions and industries; Lucidworks data from this month showed that just 20 percent of AI deployments in manufacturing have been considered successful.
Gamazaychikov said those results are promising, particularly for some of the newer additions to the AI market.
“With this technology, there’s a lot of experiments and pilots, so it was heartening to see that one-fifth of folks that are already feeling like they have this in their regular processes,” he said.
In addition to those who said AI is already fully implemented into their sustainability programs, 29 percent said their teams continue to experiment with AI for sustainability.
Sixty-five percent of those using AI in sustainability initiatives said the technology has proven transformational for their key initiatives and programs. Half of respondents said their organizations use AI to improve energy efficiency, while 48 percent said they use it for carbon emissions modeling and 47 percent indicated they use it to ensure compliance with regulations and standards.
Gamazaychikov said those types of use cases represent a strong example of the way AI can remove some of people’s grunt work, particularly on teams where big-picture strategy and movement matters deeply.
“Sustainability professionals these days are spending a lot of time doing reporting and reading through regulations, that kind of thing. That’s not necessarily the best use of everyone’s time, so it’s been great to see that these types of tools can really automate a lot of that work and let folks focus on what’s really important,” he said.
Salesforce itself has been using AI to aggregate millions of internal data points on carbon emissions, from data center infrastructure, to supply chain, to travel. Earlier this year, the company called on governments to consider environmental ramifications when they categorize the risk of AI models.
Gamazaychikov said that AI may soon reach an inflection point where the technology gets integrated into an increasing number of processes in a reliable way. But to get there, he said, technology providers and companies in other industries have to be willing to work together—fashion and apparel companies have certainly heard that same call to action for material innovation and sustainability, but it may be time to extend that attitude to technology.
“There’s a lot of competition on how you develop the models, and there’s a lot of implications there, but we really need to come forward, get out of these black boxes and work together, because this is our mutual future. It’s our planet. These are not competitive issues; we can work together, disclose this information and drive innovation towards a more sustainable future together,” he said.
Add Comment