How firms are using AI to cut their carbon emissions

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.