Advertisers grapple with the weight of AI’s environmental impact

Generative AI has become an inescapable term in marketing. However, experts predict the environmental impact of Gen AI will be inescapable as well.

Agencies and in-house marketing teams are adopting AI across their workflows, whether to enhance targeting, optimise ad spend or develop campaigns quicker. Use of AI promises not just efficiency, but greater effectiveness in reaching audiences. 

Ninety-one percent of respondents to Forrester and the 4A’s Gen AI Survey Report noted that their agencies are either looking at AI implementation or actively utilising it in strategy development and campaign optimisation, as well as for routine functions like transcribing notes. Agencies are using an array of AI tools, with Midjourney, OpenAI, Adobe and Google's Gemini among most common according to Jeremy Lockhorn, SVP of creative technologies and innovation at the 4A's.

Brian O'Kelley, CEO of carbon measurement ad tech firm Scope 3, pointed out the competitive pressure driving AI adoption. 

“If you're a CEO of any company, you're like, ‘Wow, I could take my incredibly talented employees and make them 20% more effective.’ That's a competitive advantage,” he said. “There's a bit of excitement and a bit of FOMO [fear of missing out] happening.” 

But while AI offers transformative capabilities and tantalising efficiencies for businesses, its unchecked growth risks undermining broader corporate environmental and sustainability goals.

According to a report by the MIT Climate & Sustainability Consortium, the rapid rate of AI implementation across industries “runs counter to the massive efficiency gains that are needed to achieve net-zero greenhouse gas emissions by 2050.”

Google’s early June sustainability report showed a greenhouse gas emissions rate for 2023 that was nearly 50% higher than in 2019, largely due to AI development. The report stated, "reducing emissions may be challenging due to increasing energy demands from the greater intensity of AI compute." 

The speed of AI development and implementation across businesses is not just outpacing the tech’s profitability; it’s also moving faster than the development of adequate ecofootprint measurement tools. 

This creates a conundrum for agencies, marketers and other industry players who are racing to keep up with AI innovation while staying on course and committed to their sustainability goals.

Addressing AI’s environmental impact

Training complex AI models requires substantial computational power, often leading to increased energy consumption by data centers. According to the MIT report, the data center power requirement in North America nearly doubled from 2022 to 2023—out pacing “the demand of over 70% of countries.” 

These centers not only contribute significantly to greenhouse gas emissions, but also require cooling systems that lead to excessive water usage.

As the environmental impacts of AI become clearer, the marketing and advertising industry is beginning to grapple with the challenge. 

“The environmental thing is there. I don't think anybody is ignoring it,” said Lockhorn. 

Freddie Liversidge, VP of global media at HP, noted that his team is asking questions about how their AI usage will impact corporate sustainability goals as they notice processes “becoming more heavily reliant on server capacity and data processing.” 

But because the environmental impact of media and advertising is often not factored into larger corporate sustainability plans, his team lacks the tools they need to adequately assess that impact. 

"At HP, the primary focus is on the emissions of our printers and computers. A lot of our focus is on fixing a lot of recycled materials there, and offsetting the carbon generated by printing,” he said. Therefore, efforts to measure the environmental impact of the media team’s AI usage has been largely spearheaded by the team itself.

Most agency groups are just beginning to create policies around AI, and while some address sustainability, they lack specifics. WPP's 2023 sustainability report, for instance, states that it “aims to consider carbon reduction as we develop AI products and services” but lacks specifics on how this impact will be measured. 

Beyond clear policies, advertising and media practitioners often lack the tools to measure carbon impact. Liversidge noted that different sustainability measurement tools often provide incomplete or conflicting information.

"I could find you four or five different companies that measure media carbon footprint, and I can guarantee you not one of them will give me the same number," he said. “There's just so much nuance that different people take different approaches. It becomes very hard for us to make decisions”

Nate Gabriel, an assistant professor at Rutgers University, added that as a concept, ecofootprints are “too rough a measure to be very meaningful.” Calculating the environmental impact of corporations is often circumstantial and more complex than simply implementing a tool he said. 

According to Amy Williams, CEO of ad tech platform Good Loop, some players could be hesitant to address the carbon impact of online advertising as a whole as it “shines a light on issues we've had for a long, long time—issues of MFA, issues of fraud, issues of wasteful supply chains.”

Other executives might be put off by the perceived extra cost of measuring the sustainability impact associated with AI, said Ben Riley, general manager of North America at Seen This. 

“Their inaction tells you that because they think that there's a cost associated, they're going to get less out of it,” he said.  “We have had conversations with CEOs at agencies who say, ‘Yeah, we hear about this all the time, but I really don't care.’”

Williams added that properly researching and communicating about the topic is paramount in order to avoid greenwashing and inflated positive results. “Those things really undermine the movement,” she said.

What the industry can do today

While specific tools and carbon calculators may not be easily accessible today, early adopters can still be proactive about addressing the tension between AI and sustainability.

According to Lockhorn, that starts with “picking models wisely.” 

“Do you need an LLM, billions of parameters, or can you use a smaller, specialised model that uses less energy during its training process?” he said, adding that there is no “one-size-fits-all when it comes to gen AI.”

Experts also emphasised the importance of establishing industry standards for sustainability measurements to get to a place where AI’s impact can be accurately calculated.

“We need standardisation around carbon footprint measurement in media first. Then we can start to move to the next step,” said Liversidge. 

Both Williams and O’Kelley pointed to the framework released by the Global Alliance for Responsible Media (GARM) in June as a starting point. According to O’Kelley, the framework, created with cross-industry input, helps ensure that self-reported sustainability measures are accurate.  

“We can actually trust that it's the right number, both from an environmental perspective, but also from a media perspective,” he said. 

While AI’s carbon footprint specifically is not directly addressed in the GARM framework, Williams said it inevitably will be and believes “that's something GARM has planned for.” 

The need for regulation

Several experts also pointed out the need for government regulation beyond industry standards maintained in good faith.

Williams stated that until we have a “GDPR for the environment,” industry standards will have to rely on moral upstanders. “By bringing regulation in, we start to see the hygiene factors increase, everybody's standards rise and that kind of rising tide should lift all ships,” she said. 

Riley said he can see a future in which late adopters will be forced to examine the footprint of AI prior to implementation, due to potential future government policy.  

“My warning to them is you need to get ahead of this in order to protect your own business, because the marketers who are demanding it of their agency partners will say, ‘we'll take our business to this one that's more aligned with our corporate goals,’” he said. 

While the conversation surrounding AI’s environmental impact is lagging behind the rapid-fire pace of implementation, addressing the issue will eventually be unavoidable. 

“People are realising that programmatic media, digital media, is some of the worst when it comes to carbon footprint,” said Liversidge. “AI sits on the majority of that media. It won't be long until we start connecting the dots.”