How firms are using AI to cut their carbon emissions
AI is increasingly being recognised as an important tool for businesses in the pursuit of their net-zero targets – although its own energy requirements are not inconsiderable
There has been a growing realisation among businesses in recent years that becoming environmentally sustainable is a must, not a choice. Customers, investors and employees and industry regulators are all putting pressure on them to act before the climate crisis worsens to calamitous levels.
Alongside this, the willingness of companies to publicise their progress in reducing their ecological impact is increasing. More than 3,400 organisations, with a combined market cap of £21.4tn, have registered their support for the Task Force on Climate-Related Financial Disclosures since it published its first reporting recommendations in 2017, for instance.
AI has a key role to play in helping firms to hit the ambitious net-zero CO2 emissions targets they are setting themselves. The Global AI Adoption Index 2022, IBM’s latest annual survey of uptake, found that two-thirds of the 7,500 IT chiefs it polled were either using AI to achieve sustainability goals or planning to do so.
Among other things, they’re looking to the technology to automate data collection and help them provide verifiable information on their firms’ environmental performance. But, according to the research, by far the most popular use for AI in this field is to make operations more efficient and, by extension, more environmentally friendly.
“When it comes to the successful adoption of AI, it’s crucial for companies to identify a clear challenge that it’s well suited to address, then focus on implementing the tech in a way that fits with their work streams,” says Dr Kareem Yusuf, general manager of IBM Sustainability Software and leader of its AI applications business. “For example, you could use AI to capture a product’s carbon footprint data across complex supply chains more easily and feed that into your sourcing and procurement decisions.”
It was a supply chain problem that first prompted Moosejaw, a Canadian retailer of outdoor recreation clothing and equipment, to turn to AI for a solution. Given the nature of its products, the firm was particularly interested in showing that it was constantly reviewing its environmental performance.
Moosejaw identified that many of its customers were practising a common online behaviour known as size sampling. This occurs when a consumer orders the same garment in more than one size with the intention of trying them all on at home and sending back those that don’t fit. The task of dealing with these unwanted items when they enter the reverse supply chain generates a lot of unnecessary greenhouse gas emissions.
Research published by tech firm Optoro has projected that handling returns in the US alone could be responsible for putting 23 million tonnes of CO2 into the atmosphere in 2025.
With the help of AI, Moosejaw identified that almost 15% of online purchases returned by its customers could be attributed to size sampling. It set about minimising this behaviour by using a data-driven personalisation platform called True Fit.
Sarah Curran-Usher, managing director of True Fit in EMEA, explains how the system works: “When a shopper places multiple sizes of the same item into their shopping cart, a change in the user experience prompts them to create a True Fit profile. The platform can then pair that consumer’s data with data in its fashion genome to recommend the best fit. Using this AI, Moosejaw has been able to reduce the rate of size sampling by 24% in one year.”
Whether it’s deployed at the customer interface or on a purely operational level, AI can help firms to extract precious gems of insight from the mountain of data they’re sitting on. But there are barriers. For instance, O’Reilly Media’s AI Adoption in the Enterprise 2022 research report has highlighted a potentially problematic lack of governance. The study revealed that more than half of the organisations it polled that were using AI didn’t have a governance plan in place. It found that the biggest problem was a lack of in-house expertise to get at the necessary data.
The race for the data to help organisations reduce their emissions and strive for carbon neutrality also appears to be fuelling an unprecedented acceleration in the uptake of AI. Tortoise Media’s Global AI Index research has tracked a significant upsurge in adoption over the past two years. Half of the firms in its most recent survey revealed that they had recently stepped up their AI usage.
“We’ve reached a point in AI maturity where we’ve identified a multitude of practical applications and decision points where AI can make a meaningful impact,” Yusuf says. “For example, companies can look to automatically link their energy usage with their physical asset systems to identify predictive maintenance opportunities to improve their environmental performance. Or, when it comes to the operation of data centres, firms could use such insights to optimise workloads and schedules based on their energy consumption.”
The operation of data centres highlights a dichotomy at the heart of the relationship between AI and sustainability. For all the enthusiasm that firms have shown for the technology’s environmental applications, this must be tempered by the knowledge that using AI itself consumes significant amounts of energy.
Nowhere is that more evident than in the humming global data centres of Google, a company that’s responsible for a significant proportion of the world’s electricity output, particularly the energy needed to keep its servers cool and thereby reduce the risk of a devastating outage.
Since September 2020, Google has been working to achieve its pledge of running entirely on carbon-free energy by 2030. It’s been applying AI to this challenge, using it to predict the combined effects of various procedures in its data centres on energy consumption and identifying where efficiencies can be achieved.
Professor Ong Yew Soon, chief AI scientist at Singapore’s Agency for Science, Technology and Research, believes that there is further potential to cut the energy consumption of AI systems.
“Usage can be saved at the ‘training phase’, so we are looking to reduce the need for massive amounts of training data,” he says. “So, instead of training an AI system on a massive web of information, researchers can pick and choose according to the system’s application, which would save tonnes of CO2.”
It’s just another example of the link between AI and sustainability. If the technology can find a way to minimise its own carbon footprint, it will close a virtuous circle, helping its users to hit the net-zero targets that matter so much to all stakeholders in the climate crisis.