Summary: Working with AI: Real Stories of Human-Machine Collaboration

The present state of work has already advanced. Indus­tries span­ning from pub­lic trans­porta­tion to mar­ket­ing are present­ly uti­liz­ing AI. In this reveal­ing book, IT spe­cial­ists Steven Miller and Thomas H. Dav­en­port explore instances of suc­cess­ful AI inte­gra­tion in the work­place. They move beyond mere num­bers to pro­vide a detailed view of the cur­rent impacts of AI on work pat­terns and raise con­cerns about the future. Whether you are anx­ious about how AI affects your job or con­tem­plat­ing lead­ing AI imple­men­ta­tion in your insti­tu­tion, you will dis­cov­er valu­able per­spec­tives in Miller and Dav­en­port’s work.

Main Points

• AI automa­tion encom­pass­es rules-based frame­works along with machine learn­ing systems.
• The bulk of spon­sors inter­viewed endorsed a mod­el of AI adop­tion that boosts work­ers’ effi­cien­cy rather than dis­plac­ing them.
• AI is fos­ter­ing deep­er align­ment between busi­ness and IT oper­a­tions, lead­ing to a need for hybrid roles.
• The in-depth under­stand­ing of their roles by front­line work­ers is cru­cial for suc­cess­ful AI integration.
• AI is pro­duc­ing a vari­ety of effects on the demand for begin­ner-lev­el employees.
• There are numer­ous tasks machines are still unable to handle.

Overview

AI automation encompasses rules-based systems as well as machine learning frameworks.

AI devel­op­ers con­struct rules-based sys­tems using mul­ti­ple if-then con­di­tions. They edu­cate machine learn­ing frame­works on labeled data. These frame­works can iden­ti­fy pat­terns in this data, which they sub­se­quent­ly uti­lize to ana­lyze unla­beled data. In pro­fes­sion­al set­tings, AI serves var­i­ous pur­pos­es, including:

• Pre­dic­tions
• Rec­om­men­da­tions
• Rank­ings (e.g., for sales leads)
• Locat­ing spe­cif­ic details in a document
• Automat­ing processes

“Numer­ous indi­vid­u­als cur­rent­ly inter­act with AI on a dai­ly basis. We observed this phe­nom­e­non in large cor­po­ra­tions and small busi­ness­es, in offices, fac­to­ries, farms, and across a wide spec­trum of intel­lec­tu­al and admin­is­tra­tive tasks.”

The inte­gra­tion of AI for automa­tion and enhance­ment in work­places is already preva­lent and vis­i­ble in many cur­rent work envi­ron­ments. At Mor­gan Stan­ley, AI offers tai­lored sug­ges­tions for prod­ucts and pro­grams based on a clien­t’s indi­vid­ual pro­file and port­fo­lio. The finan­cial advi­sor assess­es the rel­e­vance of these sug­ges­tions based on their per­son­al famil­iar­i­ty with the client. This facil­i­tates more effec­tive client com­mu­ni­ca­tion, as AI can cus­tomize rec­om­men­da­tions for each client to a lev­el that human advi­sors would find time-con­sum­ing. Advi­sors note that, with AI’s assis­tance, they can pro­vide more per­son­al­ized guid­ance to a larg­er clientele.

At Jew­el Mall in Sin­ga­pore, an AI sys­tem now par­tial­ly auto­mates the inci­dent report fil­ing process for the secu­ri­ty team. The AI pre-pop­u­lates reports with per­ti­nent data and sim­pli­fies the inclu­sion of addi­tion­al infor­ma­tion such as pho­tos and video from the CCTV network.

The majority of sponsors interviewed endorsed a model of AI adoption that boosts workers’ efficiency rather than displacing them.

The spon­sors of an AI imple­men­ta­tion ini­tia­tive are the pri­ma­ry dri­vers in today’s work envi­ron­ment. Nor­mal­ly, senior man­agers define the direc­tion that influ­ences mod­i­fi­ca­tions to busi­ness process­es and spon­sor new train­ing for employees.

Automa­tion can have two broad impacts on jobs: either sub­sti­tut­ing human roles through com­plete automa­tion or enhanc­ing human work via par­tial automa­tion. Thus far, enhance­ment has been the pre­vail­ing trend. In 2018, Deloitte car­ried out a sur­vey of US exec­u­tives. 63% of those famil­iar with AI indi­cat­ed they would replace work­ers with machines for cost-sav­ing pur­pos­es, yet this was not the exec­u­tives’ prin­ci­pal motive for embrac­ing AI. Exec­u­tives were more inter­est­ed in enhanc­ing prod­ucts or inter­nal process­es, aid­ing in deci­sion-mak­ing, and free­ing work­ers to con­cen­trate on cre­ative duties.

“If you’re trou­bled about the impli­ca­tions of AI on employ­ment, it’s heart­en­ing that indi­vid­u­als in var­i­ous roles are nec­es­sary to con­struct, imple­ment, oper­ate, and sus­tain these systems.”

DBS Bank, the largest bank in South­east Asia, inte­grat­ed AI to assist their anti-mon­ey-laun­der­ing ana­lysts. The AI is sup­port­ed by more rou­tine aspects of analy­sis, enabling ana­lysts to bet­ter exam­ine emerg­ing risks. Rather than striv­ing for the ulti­mate sub­sti­tu­tion of employ­ees by AI, the exec­u­tive lead­ing the change at DBS aimed to aid human work­ers in per­form­ing their roles more effec­tive­ly. This objec­tive fos­tered a sup­port­ive envi­ron­ment for employ­ees col­lab­o­rat­ing with AI, moti­vat­ing them to offer feed­back, share exper­tise, and col­lab­o­rate with AI.

Major economies are grap­pling with an aging pop­u­la­tion and a man­pow­er short­age, hence com­pa­nies are like­ly to increas­ing­ly resort to automa­tion to com­pen­sate for the defi­cien­cy in labor, rather than replac­ing employ­ees. While the total work­force required by firms using enhance­ment may be lean­er, in many com­pa­nies sur­veyed, the work­force did not con­tract and even con­tin­ued expand­ing due to busi­ness growth – itself pro­pelled by AI-dri­ven pro­duc­tiv­i­ty enhancements.

The suc­cess­ful exe­cu­tion of AI sys­tems hinges not only on new tech­nol­o­gy but also on new busi­ness mod­els, process­es, and work­er com­pe­ten­cies. While AI will sup­plant or enhance numer­ous posi­tions, it also gen­er­ates new roles involved in strate­giz­ing, craft­ing, imple­ment­ing, super­vis­ing, and enhanc­ing AI systems.

AI is fostering deeper alignment between business and IT operations, leading to a demand for hybrid roles.

Tra­di­tion­al­ly, indi­vid­u­als in busi­ness and IT func­tions have had lim­it­ed com­pre­hen­sion of each oth­er’s respon­si­bil­i­ties. Busi­ness func­tions encom­pass areas like human resources, mar­ket­ing, finance, and man­age­ment. IT func­tions cov­er the estab­lish­ment or con­fig­u­ra­tion of the IT sys­tems used by indi­vid­u­als, as well as data-focused roles such as data sci­en­tists, ana­lyt­ics experts, AI/­ma­chine-learn­ing engi­neers, and data engi­neers. Cur­rent­ly, a plau­si­ble gap in knowl­edge exists between AI teams and oth­er IT professionals.

Mul­ti-func­tion­al roles bridge the gap between IT and busi­ness knowl­edge. Some­times this bridg­ing involves over­ar­ch­ing coor­di­na­tion, like the prod­uct man­ag­er for AI sys­tems and ser­vices at Shopee, an e‑commerce site in South­east Asia, who ensures coher­ence through­out the orga­ni­za­tion and final­izes deci­sions among mul­ti­ple stake­hold­ers. In oth­er cas­es, it entails mul­ti­dis­ci­pli­nary squads con­cen­trat­ing more specif­i­cal­ly on gov­er­nance, com­pli­ance, or ethics. The Sales­force eth­i­cal AI prac­tices team under­takes out­reach endeav­ors to stim­u­late indi­vid­u­als to con­tem­plate the eth­i­cal impli­ca­tions of their choic­es and assists oth­er teams in resolv­ing par­tic­u­lar eth­i­cal inquiries, such as evad­ing bias in the datasets used to train AI.

The neces­si­ty for these spe­cial­ized roles to act as a con­nec­tor between two pre­dom­i­nant­ly dis­tinct knowl­edge domains only empha­sizes the dis­par­i­ty that endures between most indi­vid­u­als in IT and those in busi­ness roles. Nonethe­less, indi­ca­tors sug­gest this sit­u­a­tion is start­ing to change. Present­ly, orga­ni­za­tions incor­po­rat­ing AI usu­al­ly have at least one indi­vid­ual in a busi­ness role who focus­es inten­sive­ly on address­ing busi­ness chal­lenges with data and tech­nol­o­gy. Despite their pri­ma­ry exper­tise lying in busi­ness, these pro­fes­sion­als have acquired an ade­quate under­stand­ing of the tech­nol­o­gy they engage with to engage in dia­logues with IT pro­fes­sion­als. This trend is antic­i­pat­ed to per­sist as it emerges as a log­i­cal con­se­quence of the infil­tra­tion of IT pro­ce­dures into all facets of business.

“Numer­ous indi­vid­u­als in IT divi­sions now pos­sess busi­ness back­grounds instead of tech­no­log­i­cal ones.”

More­over, count­less enter­pris­es now antic­i­pate their IT experts to pos­sess a sol­id under­stand­ing of the busi­ness chal­lenges they are striv­ing to resolve. For instance, the online styling ser­vice Stitch Fix man­dates that their data sci­en­tists acquire the skill of styling clients. A pro­found com­pre­hen­sion of how styl­ists uti­lize IT’s tech­ni­cal solu­tions enables the head of data sci­ence to more effec­tive­ly assess the impacts of her cod­ing selections.

Often­times, dri­ven indi­vid­u­als iden­ti­fy an oppor­tu­ni­ty to merge an IT or AI solu­tion in their estab­lish­ment and acquire the nec­es­sary exper­tise. Jen­nifer Schmich at Intu­it, a provider of finan­cial soft­ware, com­menced her career as a copy­writer before rec­og­niz­ing an open­ing to begin uti­liz­ing Writer.com, an AI-pow­ered writ­ing assis­tant. Schmich col­lab­o­rat­ed with her super­vi­sor to craft her posi­tion as a con­tent archi­tect and now super­vis­es a small team capa­ble of coor­di­nat­ing reg­u­la­tions such as style guides and stan­dard­ized lan­guage for thou­sands of writers.

Com­pet­i­tive busi­ness­es must embrace data. This sig­ni­fies that the quan­ti­ty of hybrid IT-busi­ness roles will only esca­late. Far­sight­ed com­pa­nies insist that IT pro­fes­sion­als gain expo­sure to the oper­a­tions of their busi­ness units and vice versa.

The profound knowledge of their roles by frontline workers is crucial for the successful integration of AI

The pro­fes­sion­al exper­tise of front­line work­ers is vital in effec­tive­ly inte­grat­ing AI into their tasks. They are fre­quent­ly called upon to assess AI sug­ges­tions or out­puts, there­by empha­siz­ing the impor­tance of their pro­fes­sion­al judg­ment. There­fore, train­ing employ­ees who lack exten­sive expe­ri­ence in their roles to col­lab­o­rate with AI could present a challenge.

Front­line work­ers inter­viewed view their role as essen­tial for eval­u­at­ing machine rec­om­men­da­tions, incor­po­rat­ing strate­gic think­ing, and col­lab­o­rat­ing with oth­ers. Many men­tioned that AI less­ened monot­o­ny and made their work more men­tal­ly stim­u­lat­ing. Nev­er­the­less, some found that the height­ened focus on com­mu­ni­ca­tion and intel­lec­tu­al work increased the demands of their roles. In every case study con­duct­ed for this book, employ­ee effi­cien­cy improved with AI integration.

“Mov­ing for­ward, ini­tia­tives regard­ing sys­tem design, imple­men­ta­tion, and ongo­ing post-deploy­ment oper­a­tional assis­tance will con­tin­ue to yield bet­ter out­comes with par­tic­i­pa­to­ry con­tri­bu­tions and robust back­ing from front­line workers.”

Assign­ing indi­vid­u­als from var­i­ous sec­tors with­in the orga­ni­za­tion not only to adopt auto­mat­ed solu­tions but also to aid in design­ing them per­mits a nuanced approach. The robot­ic process automa­tion team at Japan­ese inter­na­tion­al adver­tis­ing and pub­lic rela­tions agency Dentsu dis­cov­ered that the com­pa­ny, akin to numer­ous knowl­edge-inten­sive work­places, did not rely on broad, read­i­ly auto­mat­ed busi­ness process­es. Instead, the team iden­ti­fied an exten­sive list of micro-tasks spe­cif­ic to indi­vid­u­als’ work, prompt­ing them to engage employ­ees from across the orga­ni­za­tion. They fur­nished these employ­ees with train­ing and a tool that enabled them to devise their own automa­tion rou­tines for their repet­i­tive duties. This ini­tia­tive saved approx­i­mate­ly 3500 work hours.

AI is exerting varied impacts on the demand for entry-level employees

In cer­tain indus­tries, the trend is toward hir­ing few­er entry-lev­el work­ers. This cir­cum­stance is evi­dent in rou­tine phys­i­cal tasks like burg­er flip­ping or weed­ing, as well as clear­ly defined, rou­tine men­tal tasks such as visu­al qual­i­ty inspec­tions. As a result of AI inte­gra­tion, Haven Life/MassMutual present­ly requires few­er entry-lev­el insur­ance under­writ­ers as they tend to auto­mate tasks. How­ev­er, they still neces­si­tate expe­ri­enced under­writ­ers for their abil­i­ty to eval­u­ate the AI sys­tem’s out­put and han­dle non-rou­tine cases.

Con­verse­ly, AI-pow­ered train­ing can actu­al­ly low­er the bar­ri­er to entry for cer­tain roles. For instance, the PBC Lin­ear machine shop employs an AI-sup­port­ed aug­ment­ed real­i­ty train­ing sys­tem named Taqtile to expe­dite the train­ing of man­u­fac­tur­ing work­ers. Tac­tile oper­ates with an aug­ment­ed real­i­ty plat­form uti­liz­ing AI to adapt to each user, achiev­ing a pre­ci­sion lev­el imper­a­tive for machine-shop train­ing. The sys­tem deliv­ers per­son­al­ized train­ing that would have for­mer­ly man­dat­ed one-on-one coach­ing, allow­ing new hires to learn at their own pace.

“The very same tech­nol­o­gy… can both dimin­ish oppor­tu­ni­ties for entry-lev­el work­ers through effi­cien­cy enhance­ments and broad­en entry-lev­el employ­ment oppor­tu­ni­ties via height­ened lev­els of embed­ded train­ing, guid­ance, and per­for­mance support.”

AI sys­tems can also aid in enhanc­ing job access for groups encoun­ter­ing capa­bil­i­ty chal­lenges. Dentsu col­lab­o­rat­ed with Auton­o­my­Works, an orga­ni­za­tion spe­cial­iz­ing in gen­er­at­ing job prospects for indi­vid­u­als on the autism spec­trum, to cre­ate spe­cial­ized work tools for autis­tic employees.

The dimin­ish­ing quan­ti­ty of entry-lev­el job prospects pos­es a sys­temic issue for the econ­o­my and soci­ety. While the impact of wide­spread AI adop­tion remains to be seen, the reduc­tion of entry-lev­el posi­tions in knowl­edge-based sec­tors pos­es a sig­nif­i­cant threat.

There remain numerous tasks machines are unable to perform.

Human judg­ment remains a fun­da­men­tal aspect in most work aug­ment­ed by machine learning:

• AI can­not com­pre­hend con­text – Devel­op­ers are unable to encap­su­late con­text with­in datasets uti­lized to train machine-learn­ing algo­rithms, nor can they encap­su­late such a broad con­cept with­in algo­rith­mic reg­u­la­tions. Con­se­quent­ly, AI is inca­pable of uti­liz­ing data to con­struct a coher­ent nar­ra­tive, define a prob­lem, ren­der sub­jec­tive assess­ments, or con­tem­plate the broad­er social and eth­i­cal impli­ca­tions of its actions.
• Com­plex sys­tems also present a hur­dle for AI – Sev­er­al human sys­tems are exces­sive­ly intri­cate for AI to deci­pher. As a result, AI can­not dif­fer­en­ti­ate cru­cial alerts from insignif­i­cant ones in com­plex set­tings, such as a pub­lic space under­go­ing secu­ri­ty sur­veil­lance. It is inept at nego­ti­at­ing or align­ing deci­sions among groups with vary­ing or evolv­ing pri­or­i­ties, nor can it con­vince indi­vid­u­als to adopt new behav­iors or elim­i­nate orga­ni­za­tion­al bar­ri­ers to effect orga­ni­za­tion­al change.
• AI is still inca­pable of inter­pret­ing emo­tion­al con­texts– Despite pop­u­lar por­tray­als of AI sys­tems cul­ti­vat­ing rela­tion­ships with humans, akin to those depict­ed in films like Her and Ex Machi­na, AI is inca­pable of under­stand­ing emo­tion­al needs, form­ing rela­tion­ships with humans, enhanc­ing job sat­is­fac­tion or employ­ee morale, or dis­sect­ing the tonal­i­ty of writ­ten com­mu­ni­ca­tion. Even the Writer AI tool encoun­ters chal­lenges in ana­lyz­ing tone, and AI sys­tems employed for social media ana­lyt­ics strug­gle to detect sarcasm.
• AI is reliant on human sup­port – AI still man­dates human inter­ven­tion to con­fig­ure the phys­i­cal sys­tems or envi­ron­ments nec­es­sary for gath­er­ing data for analy­sis, rec­ti­fy­ing mal­func­tions in AI, and trans­fer­ring exper­tise from human spe­cial­ists to AI systems.

For all these rea­sons, humans must still make the final deci­sions con­cern­ing AI-gen­er­at­ed recommendations.

“One of the prin­ci­pal advan­tages of humans and intel­li­gent machines col­lab­o­rat­ing is that humans can val­i­date that an auto­mat­ed deci­sion is ‘sen­si­ble.’”

It is cru­cial for humans col­lab­o­rat­ing with machines to com­pre­hend the busi­ness pro­ce­dures that AI is trained to facil­i­tate, enabling them to grasp the ratio­nale behind AI deci­sions and eval­u­ate their ade­qua­cy. Like­wise, AI sys­tems neces­si­tate fea­tures that pro­vide expla­na­tions for the deci­sions they make to human work­ers, empow­er­ing humans to obtain the nec­es­sary details to assess those choic­es. As an illus­tra­tion, DBS Bank’s anti-mon­ey laun­der­ing sys­tem includes a pan­el that elu­ci­dates the risk scores gen­er­at­ed by the AI. Clar­i­fy­ing AI judg­ments also moti­vates employ­ees to embrace AI inte­gra­tion with­in their positions.

Regarding the Authors

Thomas H. Dav­en­port serves as a Pro­fes­sor of Infor­ma­tion Tech­nol­o­gy and Man­age­ment at Bab­son Col­lege and acts as a senior AI con­sul­tant for Deloitte Ana­lyt­ics. Steven Miller is a Pro­fes­sor Emer­i­tus of Infor­ma­tion Sys­tems at Sin­ga­pore Man­age­ment University.

Evaluation

“Work­ing with AI: Real Sto­ries of Human-Machine Col­lab­o­ra­tion” by Steven M. Miller and Thomas H. Dav­en­port offers an exten­sive and per­cep­tive explo­ration of the prag­mat­ic uses of arti­fi­cial intel­li­gence (AI) across diverse sec­tors. It presents a com­pi­la­tion of authen­tic anec­dotes and case stud­ies that show­case the coop­er­a­tive part­ner­ship between humans and AI systems.

The com­mence­ment of the book sets a firm ground­work for com­pre­hend­ing AI and its abil­i­ties. It delin­eates the var­i­ous cat­e­gories of AI tech­nolo­gies and their poten­tial influ­ence on the labor force and soci­ety. The authors under­score the sig­nif­i­cance of accept­ing AI as a tool for enhance­ment rather than sub­sti­tu­tion, advo­cat­ing the notion that humans and machines can col­lab­o­rate to achieve supe­ri­or results.

Through­out the book, Miller and Dav­en­port share myr­i­ad instances of tri­umphant human-machine col­lab­o­ra­tions in var­ied domains. They exhib­it how AI has been assim­i­lat­ed into realms such as health­care, finance, man­u­fac­tur­ing, and cus­tomer ser­vice to boost pro­duc­tiv­i­ty, refine deci­sion-mak­ing, and fos­ter inno­va­tion. The authors shed light on the hur­dles encoun­tered dur­ing these incor­po­ra­tions and fur­nish valu­able insights into how estab­lish­ments can nav­i­gate the com­plex­i­ties of AI adoption.

Key Points:

• AI is now a con­crete actu­al­i­ty that is reshap­ing busi­ness­es and sectors.
• The tri­umph of AI inte­gra­tion hinges on the capac­i­ty of humans and machines to col­lab­o­rate proficiently.
• AI can ampli­fy human capa­bil­i­ties, yet it can­not entire­ly sup­plant human judg­ment and decision-making.
• The forth­com­ing work land­scape will entail a fusion of human and machine col­lab­o­ra­tion, with AI man­ag­ing monot­o­nous and rou­tine tasks while humans con­cen­trate on high-lev­el deci­sion-mak­ing and inven­tive work.

One of the strong suits of “Work­ing with AI” lies in its focus on the human facet in AI deploy­ments. The authors under­score the neces­si­ty for effec­tive change man­age­ment, address­ing appre­hen­sions regard­ing job dis­place­ment, and fos­ter­ing a cul­ture of col­lab­o­ra­tion between humans and AI sys­tems. They accen­tu­ate the impor­tance of enhanc­ing skills and retrain­ing the labor force to ful­ly exploit the poten­tial of AI technology.

Fur­ther­more, the book delves into eth­i­cal con­sid­er­a­tions asso­ci­at­ed with AI. Miller and Dav­en­port delib­er­ate on the plau­si­ble bias­es and unin­ten­tion­al con­se­quences that may arise when uti­liz­ing AI sys­tems, empha­siz­ing the need for trans­paren­cy, equi­ty, and answer­abil­i­ty. They offer prac­ti­cal coun­sel on ensur­ing eth­i­cal AI prac­tices and urge read­ers to approach AI deploy­ment with a con­sci­en­tious mindset.

The writ­ing style of the book strikes a hap­py medi­um between tech­ni­cal intri­ca­cy and acces­si­bil­i­ty. The authors expound intri­cate con­cepts clear­ly and con­cise­ly, ren­der­ing it suit­able for both tech­ni­cal and non-tech­ni­cal read­ers. The inclu­sion of real-life case stud­ies lends cred­i­bil­i­ty and allows read­ers to relate to the mate­r­i­al, show­cas­ing the poten­tial of AI in con­crete ways.

While “Work­ing with AI” encom­pass­es a broad array of sec­tors and appli­ca­tion sce­nar­ios, cer­tain read­ers may per­ceive that par­tic­u­lar domains receive more atten­tion than oth­ers. Addi­tion­al­ly, the book pre­dom­i­nant­ly con­cen­trates on pros­per­ous AI imple­men­ta­tions, and although it fleet­ing­ly men­tions chal­lenges, a more thor­ough explo­ration of poten­tial set­backs and fail­ures could have enriched the nar­ra­tive further.

Ulti­mate­ly, “Work­ing with AI” is an excel­lent­ly script­ed and enlight­en­ing book that fur­nish­es read­ers with an across-the-board com­pre­hen­sion of the real-world appli­ca­tions of AI. The book’s prac­ti­cal coun­sel and real-life anec­dotes ren­der it an indis­pens­able resource for any­one seek­ing to har­ness AI in their occu­pa­tion or busi­ness. I unequiv­o­cal­ly endorse this book to indi­vid­u­als aim­ing to acquire a pro­found under­stand­ing of the ram­i­fi­ca­tions of AI on the future of work.