The ways AI will change the future of work

Since the arri­val of ChatGPT, there has been concern that we are going to be repla­ced by robots. Gene­ra­tive arti­fi­cial intel­li­gence, capable of assi­mi­la­ting and crea­ting writ­ten, visual, or audio content, is often des­cri­bed as a threat to jobs. With each new major tech­no­lo­gi­cal advance comes its share of debate and appre­hen­sion about impacts on the work­force. During the indus­trial revo­lu­tion, manual wor­kers were at the fore­front of these major changes. Conver­se­ly, AI is now more rele­vant to mana­gers and pro­fes­sio­nals. But what will be the real impact of this technology ?

Toge­ther with Pawel Gmy­rek and David Bes­cond, Inter­na­tio­nal Labour Orga­ni­sa­tion eco­no­mist Janine Berg ana­ly­sed the 436 occu­pa­tions lis­ted in the ILO’s Inter­na­tio­nal Clas­si­fi­ca­tion of Occu­pa­tions. The aim was to unders­tand which types of jobs would be most affec­ted by AI on a glo­bal scale. The authors used ChatGPT to ana­lyse the tasks asso­cia­ted with the occu­pa­tions and assi­gned them scores cor­res­pon­ding to their poten­tial expo­sure. Some tasks are high­ly expo­sed to tech­no­lo­gy, others less so. The higher the expo­sure poten­tial of an acti­vi­ty, the more like­ly it is to be automated.

For eco­no­mists, the pri­ma­ry impact of arti­fi­cial intel­li­gence will not real­ly be the mas­sive des­truc­tion of jobs, but rather the pro­found trans­for­ma­tion of work. For most pro­fes­sions, cer­tain tasks will indeed be car­ried out by bots, but this will leave time for other, more com­plex acti­vi­ties. On ave­rage, 10–13% of jobs world­wide could be “aug­men­ted” or trans­for­med. The first jobs to use this tech­no­lo­gy will poten­tial­ly be ware­hou­se­men, deli­ve­ry dri­vers, mana­gers in the dis­tri­bu­tion sec­tor, machine ope­ra­tors and assem­blers, ser­vice and sales wor­kers, dri­ving ins­truc­tors, dri­vers, wai­ters, archi­tects, tea­chers, musi­cians, etc. In all, 427 mil­lion jobs, or 13% of jobs world­wide, could change because of arti­fi­cial intelligence.

Although the poten­tial for change is much grea­ter than auto­ma­tion, the risk remains very real, with 2.3% of jobs world­wide affec­ted. Admi­nis­tra­tive jobs would be hea­vi­ly impac­ted by auto­ma­tion. “Call centre employees, secre­ta­ries, data entry ope­ra­tors – simple, linear acti­vi­ties with lit­tle varia­tion in tasks and lit­tle inter­ac­tion with others – could be repla­ced by bots,” explains Janine Berg. In recent years, office wor­kers have alrea­dy seen their day-to-day work evolve. Accor­ding to the experts, 24% of their tasks are high­ly expo­sed to AI, and 58% are mode­ra­te­ly expo­sed. This is by far the occu­pa­tion most at risk. This means that 2.3% of jobs world­wide, or 75 mil­lion, could end up being automated.

Our stu­dy should not be read as a reas­su­ring voice, but rather as a call to deve­lop mea­sures to deal with the immi­nent tech­no­lo­gi­cal changes

Arti­fi­cial intel­li­gence will clear­ly not affect all pro­fes­sions in the same way. The tech­no­lo­gy is also like­ly to have dif­ferent conse­quences for men and women. Women will be 2.5 times more affec­ted by auto­ma­tion than men, not least because there are more women in low-skilled admi­nis­tra­tive posi­tions. Conver­se­ly, male-domi­na­ted pro­fes­sions such as secu­ri­ty, trans­port and construc­tion are unli­ke­ly to be affec­ted. As a result, 3.7% of women’s jobs world­wide are at risk of being auto­ma­ted, com­pa­red with 1.4% of men’s jobs. This dif­fe­rence is even grea­ter in rich coun­tries, with 7.8% of jobs held by women like­ly to be repla­ced by bots, com­pa­red with 2.9% of jobs held by men. In low-income coun­tries, fewer women are in the labour mar­ket, and low-skilled admi­nis­tra­tive pro­fes­sions are pre­do­mi­nant­ly held by men.

The other major dif­fe­rence poin­ted out by the ILO eco­no­mists depends on the wealth of the coun­tries concer­ned. “In low-income coun­tries, arti­fi­cial intel­li­gence is unli­ke­ly to be deployed. The tech­no­lo­gy is expen­sive, and there is a lack of infra­struc­ture, with a poor elec­tri­ci­ty sup­ply and poor inter­net connec­tion”, explains Janine Berg. In fact, by 2022, a third of the world’s popu­la­tion will not have access to the inter­net. What’s more, the struc­ture of the labour mar­ket in low-income coun­tries makes them less sus­cep­tible to auto­ma­tion. In these coun­tries, 0.4% of jobs could be repla­ced by bots, com­pa­red with 5.5% in high-income coun­tries. In terms of pos­sible changes to jobs, 10.4% of occu­pa­tions are affec­ted in low-income coun­tries, com­pa­red with 13.4% in rich coun­tries. In short, poten­tial auto­ma­tion main­ly concerns rich coun­tries. They will be more dis­rup­ted by AI, but they will also be able to take advan­tage of it. “This situa­tion could create a pro­duc­ti­vi­ty divide bet­ween rich and poor coun­tries”, says the economist.

While there are nuances depen­ding on the region of the world, the stu­dy broad­ly envi­sages the inte­gra­tion of AI into eve­ry­day life. Repla­cing humans with bots is not on the cards for the time being. “This approach could have been expec­ted to gene­rate an alar­ming num­ber of job losses, but is not the case. Our ove­rall esti­mate points more towards a future where work is in fact trans­for­med, but still present”, the eco­no­mists sum­ma­rise. Howe­ver, this evo­lu­tion of work must take cer­tain issues into account to avoid a nega­tive impact. “Our stu­dy should not be read as a reas­su­ring voice, but rather as a call to deve­lop mea­sures to deal with the immi­nent tech­no­lo­gi­cal changes”, explain the authors.

Janine Berg believes that gene­ra­tive AI is fun­da­men­tal­ly nei­ther posi­tive nor nega­tive. It all depends on how the tech­no­lo­gy is deployed. The eco­no­mist details a num­ber of actions that govern­ments need to take : “reflec­ting on the ques­tion of the balance of power, the voice of wor­kers affec­ted by labour mar­ket adjust­ments, res­pect for exis­ting stan­dards and rights, and the appro­priate use of natio­nal social pro­tec­tions, as well as trai­ning sys­tems will be cru­cial ele­ments in stee­ring the deploy­ment of AI in the world of work.”

It is not just a ques­tion of obser­ving how this new tech­no­lo­gy is applied, but of sup­por­ting it with reflec­tion and mea­sures. The aim is to ensure social dia­logue, rede­ploy­ment or trai­ning for employees affec­ted by auto­ma­tion, and employee par­ti­ci­pa­tion in the intro­duc­tion of AI for those whose tasks will be trans­for­med. “If we don’t put mea­sures in place, and these sys­tems arrive, more jobs than neces­sa­ry will be lost. Wor­king condi­tions will dete­rio­rate. There may be short-term gains for some com­pa­nies, but there will be social conse­quences,” warns Janine Berg.

Sirine Azouaouis