Is ChatGpt the new ally of managers?

Many leaders don't know how to design roles within the company. Generative AI can be effective. But you need to use it wisely.

Fangfang Zhang, researcher at the Center for Transformative Work Design at Curtin UniversitySharon K. Parker, Professor at Curtin University

Published in the issue 1 January/February 2025 Mit Sloan Management Review Italy

Today top managers are committed to increasing the involvement of collaborators anddecrease turnover, but they must face a harsh reality: widespread burnout. Forto fight it, they must offer them healthier and more meaningful working conditions. Ina survey conducted in the United States by Gallup in 2022, 40% of workers saidthat work has a negative impact on mental health and around 30% said it didoften suffer from burnout. It is not surprising that, in the USA, worker engagement is atminimum levels, for seven years now: in 2022, only 32% of workers interviewed byGallup stated that they felt involved and 17% said they were actively disengaged (Wigert andPendell, 2023). It is estimated that this lack of engagement costs businesses 7,800 billiondollars globally, equal to 11% of global gross domestic product (Ibidem).The underlying causes of disengagement and work stress often lie in the waythe organization plans the work of its people: decades of in-depth researchconsistently link poor job design to negative effects, among themmental stress, high turnover, dissatisfaction at work, decreased productivityand compromised training (Parker, 2014).Many companies are working to improve this situation, but our researchsuggests that many managers lack the understanding to planhigh quality job positions. This is where technologies can play a key roleArtificial Intelligence (AI), such as ChatGpt: filling the gaps in their knowledge andhelping them design quality positions, for the benefit of both workers andorganizations. However, it is important that those responsible first understand the pros and cons ofuse ChatGpt to design locations.Below, we analyze some ideas in this regard highlighted by our research.

The challenge for managers: going beyond banal and unrewarding positions

What factors determine the high quality of a job position? The smart model, onejob design scheme created by Sharon K. Parker, defines a position ofhigh quality as a stimulating job (variety of tasks and possibility of developing new onesskills), endowed with mastery (clarity on the role and in feedback on work), agentive(autonomy in work and participation in change), relational (social support andpositive teamwork) and tolerable (manageable working hours and reasonable levels oftime pressure), (Parker, 2022).Despite the obvious advantages of a properly designed location, inorganisations are still prevalent those developed in an approximate manner. Secondthe Gallup Great Jobs 2019 survey, only 40% of employed Americans have a job withthe smart features described above (Rothwell and Crabtree, 2019). 16% hold oneposition lacking the essential ones and 44% have a job that presents only some attributessmart satisfiers (Rothwell and Crabtree, 2019).Results that can be explained by the lack of skills in the design ofwork by today's managers. A study we recently conducted explored thehow you design work for others (Parker, Andrei, & Van den Broeck, 2019). Aiparticipants in an online simulation (Simulation 1) were asked to design a roleadministrative, transforming a part time job, composed exclusively of fourphotocopying and archiving tasks, in one full time, selecting four tasksadditional from a list of 11, among which, five repetitive photocopying tasks andstorage and six more meaningful and interesting tasks, such as welcoming visitors orhelp colleagues organize meetings.Participants were assigned a score from 0 to 4, with a higher score indicatingthe creation of a more stimulating work project. Almost half (45%) of studentsbusiness management university students and professionals or managers who work in business servicespeople tend to incorporate repetitive and monotonous work, demonstrating a gap ofknowledge when it comes to designing quality work.In a second simulation (Simulation 2), participants were asked to act as amanager and solve a job task problem, in four different scenarios. In everyscenario, the design of the proposed role was clearly poor. To solve iproblems, participants could consider adopting strategies from a list thatincluded both 'worker correction' approaches, which attributed problems toemployee, both approaches to 'correcting inadequate work design' andrecognized the poor quality in the design of the work.One of the scenarios, for example, featured a warehouse worker who couldn'trespect 50% of the deadlines set for recovering the products and delivering them toshipment, despite the frequent withdrawals of goods. Although most of the participantshad focused on choices that improved the design of work (such as“Involve the employee and his colleagues in a review to identify howtheir work could be organized better” or “Reorganize the work so that itasks should not be timed”), a surprising number of participants (40%)has chosen strategies aimed at correcting the worker (such as "Discreetly observe thethe person's behavior to see how fast they move"). These resultsvalidate our observation on the lack of skills in the design ofwork by managers and staff.

Generative AI can suggest more robust design

In the next phase of the study, we tried to understand if a tool equipped with AI,how ChatGpt could help managers do better. The answer came quickly:ChatGpt has the potential to improve decision making in the design ofwork. However, it is important for managers to understand what ChatGpt can, and cannot, do well.We used ChatGpt to make hypothetical design decisionsadministrative task foreseen by Simulation 1. We have performed task 20times, each time using a new independent session, but with the same prompt. Inall iterations, the Generative AI chose from the list provided tasks that enriched thetask, surpassing both students and professionals and managers in creating jobsinteresting and significant. ChatGpt's performance was similar to that ofwork design experts.We then instructed ChatGpt to tackle Simulation 2, dealing with the fourmanagerial management of the workforce (including the hypothetical warehouse worker mentioned above), andby choosing strategies from a list. In every scenario, ChatGpt consistently chose thestrategies aimed at correcting work planning (good strategies for well-being,motivation and employee performance) compared to those focused on correctionof the worker. As in Simulation 1, ChatGpt outperformed students, managers andprofessionals for attention to work planning aimed at solving the problems ofstaff.In Simulation 2, however, ChatGpt did not obtain the same score as the experts,justifying a certain caution. In a session dedicated to the warehouse worker who doesn'trespected the deadlines, for example, ChatGpt obtained an overall score of 4,but he chose the strategy "I would give a bonus to the employee and her colleagues when they respect theassigned times”, which ignores the unreasonable allocation of time to the role.

A great lesson: it is essential that prompts are specific

To further evaluate ChatGpt's ability to generate suggestions in a wayindependently, we performed the Simulation 2 job design taskswithout providing a list of multiple choice solutions. We ran each test anewaccounts and in new independent chats, with history disabled, so that theChatGpt's return was not affected by the information entered previously.First of all, the result was worrying. When, in the deadline scenario notrespected, we generically asked AI to "provide effective strategies to address theproblem”, ChatGpt tended to correct the worker, creating Tayloristic approachesantiquated to systems of work, such as conducting time and travel studies.He suggested, for example, solutions such as developing time management skills and aadditional training to help the warehouse employee, who we called Karen,to improve his performance (for example, “Focus on improving speedand Karen's accuracy in locating and collecting items from the warehouse"). Yesalso recommended providing feedback on performance and incentives to motivateKaren to move faster (e.g., “By connecting performance with rewards,Karen will be encouraged to focus on improving her speed and efficiency").Each of these solutions implicitly assumes that the problem is motivation orthe ability of the worker, rather than the design of the job.After realizing that such an open approach was not optimal, welater instructed ChatGpt to deal with this same scenario using promptsspecific: “Consider designing good quality work for Karen”;

“Consider health, well-being, motivation, satisfaction and meaningfulness ofKaren's work”; and “Considering Karen's Work Design.” In this way, AIprovided some solutions to improve Karen's job design, includingfollowing:

Review time allocation and workload, to set realistic goals;Involve Karen in the decision-making process and ask for her inputimprove the work;Offer supportive feedback and recognition to increase satisfaction andemployee motivation;Provide appropriate training and skills development opportunities forincrease trust and commitment.

These strategies enhance the smart characteristics of Karen's work, increasing thetolerability of requests, enhancing agency (i.e. the awareness of being able to modifythe circumstances), increasing mastery and making the work more stimulating.ChatGpt also suggested solutions such as promoting a conducive work environment,encourage support and collaboration in the group, offer opportunities for advancementcareer and promote work-life balance. All solutions in line with theresearch on work design and well-being that would make work smarter,resulting in a more balanced and productive employee.From this particular experiment with ChatGpt we learned that Generative AI canmake valuable job design decisions, but managers must provideclear instructions that prioritize results for the worker. For example, ask ChatGptto "design high-quality work", rather than generically asking to "design agood work”, can generate more relevant and effective strategies. To improve the quality ofstrategic suggestions generated by ChatGpt, managers must therefore mentionspecific, desirable goals in their applications, such as high-quality work, health,the well-being and motivation of employees and the meaningfulness of work.The good news is that managers who don't often use ChatGpt can learn towrite effective prompts through the tool itself. For example, to the question “How socan create a well-structured and clear prompt to effectively address the challenges ofjob design?”, ChatGpt provided a step-by-step guide on how to writeeffective prompts. ChatGpt advised us to identify the main problem and bespecific, providing suggestions applicable to job design such as the use ofprompts such as “Design high-quality work that improves interaction andpositive collaboration of the sales team” or “Designing work that is meaningful tofacilitate employees' acceptance of technological changes in their environmentwork”.The experiment also demonstrated that effective prompts can be revised and tested. Imanagers who struggle with this can improve their skills by using aprompt like “The following is a prompt that aims to ask ChatGpt for advice on theview of a work design expert. Please analyze and suggestenhancements to get practical, actionable advice from ChatGpt”. In this case, ChatGptwill provide more specific suggestions for better quality results.

Five lessons for managers

To address the pervasive problem of the inadequate design of work roles andits harmful effects, it is essential to adopt innovative solutions. Based on oursresearch, ChatGpt has emerged as a promising AI tool to help managersplan more balanced and productive work. But it is not a panacea, and it must be usedwith wisdom.Below, some of our advice for managers.

ChatGpt cannot replace training.Managers must be aware that iworker problems can result from poor job design and thatthe latter, in turn, can influence the well-being, motivation and significance of thework. This means that before you can use ChatGpt to design locationsbetter, managers need training in basic design conceptsof work. To learn more about the topic, see the related article How well-designedwork makes us smarter and the Smart Work Design website.

Managers need to be supported in their goal of creating well-designed work.Imanagers must be motivated to create better working positions for their collaborators.Managers themselves therefore need roles in which the creation of work tasksbalanced for employees is seen as a legitimate and important responsibility insteadto have profit as its sole objective. If, for example, managers are promoted only inbased on productivity results, they will not be very motivated to design smart work fortheir teams.

Managers themselves need balanced work.Managers needenough time to be able to pay attention to how their work is conceivedcollaborators. This means that their work must also be tolerable and notoverly stressful.

Managers must learn to use ChatGpt prompts effectively.The managersshould be encouraged to use Generative AI to contribute to design challengesof the job, but they need clear and complete instructions on how to use it properlyeffective. For example, managers need to be specific and mention goals such asemployee well-being and motivation when they ask ChatGpt for suggestions forwork design.

ChatGpt will support, not replace, people in designing better work.Adifference from simulations, real work design situations are oftencharacterized by ambiguity and complexity, with numerous variables that can influence theresults. Real situations require a deep understanding of the context, ofhuman emotions, social dynamics and political and ethical considerations, whichcan be difficult for an AI tool like ChatGpt to understand.We recommend using ChatGpt as a complement to human managers, rather thanas their replacement.

By incorporating ChatGpt suggestions along with human expertise, organizationscan promote greater staff engagement, job satisfaction and

overall performance. By achieving this synergy between managers and AI, theroad to a better future for work design and employee experience.

Scores relating to work design decisions in the different groups.When participants in Simulation 1 were asked to design a roleadministrative, 45% chose to assign the worker repetitive and monotonous tasks(score from 0 to 4, indicating the number of enriching tasks selected by the participants).When participants in Simulation 2 were asked to address four problemswork roles designed inadequately, 40% chose strategies aimed at correcting theworker rather than bad design (score from 1 to 5, which indicates to what extent theparticipants will adopt the strategies: from 1, extremely unlikely, to 5, extremelyprobable.

Table 1.Evaluations of work design decisions in different groups

When participants in Simulation 1 were asked to design a roleadministrative, 45% mainly chose to assign repetitive tasks to the workerand monotonous (score from 0 to 4, indicating the number of enriched tasks selected by theparticipants). When participants in Simulation 2 were asked to faceproblems in four poorly designed work roles, 40% chose strategies aimed atcorrect the worker rather than the bad design (score from 1 to 5, indicating theextent to which participants will adopt the strategies, where 1 is extremely unlikely and5 is extremely probable).