Lit Review · GPT-5.5

Generative AI is beginning to transform knowledge work, which involves the creation, interpretation, communication, and application of information. Large language models and other related multimodal systems affect both the lower and upper levels of work. Unlike previous forms of automation that impacted primarily routine physical or administrative work, generative AI completes tasks including text generation, coding, image creation, and the production of summaries and analyses along with the formulation of recommendations. Automating these tasks traditionally requires significant professional judgment and expertise, and for this reason, the impacts of generative AI are being studied in a variety of disciplines including information systems, labor studies, education, management, and organizational studies. The most common focus area in the early literature is the enhancement of productivity. Field studies and early generation AI tools indicate significant time savings in writing, coding, and research. This is most noticeable in customer support and preparation for decisions in the workplace. The generation of draft content and ideas, synthesis of supporting information, design for debugging, and even translation are examples of the types of work that can be automated. This develops an opportunity for knowledge work to shift focus to judgment, creativity, and interpersonal and strategic decision-making. Some studies show that current AI tools can help improve the performance of novice workers. The literature does highlight important limitations and risks. Systems of Generative AI can produce artifacts that contain untruths or extreme biases, or that are entirely fabricated, and that can be undetectably erroneous, untruths, or artifacts of extreme bias, with a high degree of articulateness. These systems also create greater demand for verification, evaluation, and expertise within a given domain. Generative AI, instead of removing the demands for critical human thinking, will probably reallocate a large area of knowledge work to oversight, the creation of prompts, quality assurance, and the assessment of ethics. There are also many concerns related to the risks of descaling, overreliance, and a loss of control over one's own creative work due to AI systems. These and other related concerns are especially pronounced in high-risk systems such as law, health care, finance, journalism, and the public service, because erroneous outcomes can have serious socially and professionally damaging consequences. There is a strong possibility that outcomes of Generative AI systems will be highly contextualized. Creativity, Intuition, and Reflection are the pillars of Knowledge work, and the boundaries of these are shaped by many different factors, including management style, workplace culture, availability of digital resources, the legal environment, and workers' skills. For some organizations, Generative AI may be an empowering and creative resource, and may be used to redesign work practices and even potentially creative AI systems. For some organizations, Generative AI may be used to pursue a highly instrumental and even creative AI systems. For many organizations, its creative and possibly deskilling effects will be highly unidirectional and will primarily be used for the pursuit of a highly instrumental and deskilling AI systems. By examining productivity, changes in knowledge and skills, adaptation of the workplace, and ethical risks, this literature review focuses on knowledge work and the changes that will be brought about by Generative AI. Examining Generative AI in the context of professional knowledge work and existing systems of automation will hopefully help clarify some of the contradictions, tensions, and risks associated with them.

Generative artificial intelligence (AI) or large language models and multimodal systems for generating text, images, and other media, are increasingly becoming part of knowledge work, which is work that is based on the creation, interpretation, communication, or application of information. Unlike the automation of routine physical or administrative work, generative AI can produce outputs in fields of work that typically require considerable expertise on the part of professionals. There is therefore considerable interest in the potential effects of generative AI in fields such as information systems, management, labor economics, education, and organization studies. The generative capabilities of AI can increase the productivity of workers who are engaged in medium complexity tasks such as writing, searching, debugging, customer service and more. For example, a first draft of a document can be created for a knowledge worker, information from multiple sources can be aggregated for him or her, ideas can be generated for him or her, and information can be translated into other formats for him or her. This tool can also reduce the performance difference between experienced and inexperienced workers because examples, prewritten text, and structures solutions can be provided to them that would otherwise take them a lot of time to develop on their own. However, there are a number of limitations with this type of content generation, not least the risk of inaccuracy, or bias – whether this is being done accidentally or with the intent of misleading others. The fact that it is very difficult to verify as it is extremely high quality and therefore very plausible, and requires a huge amount of time to verify both in terms of reading and in terms of checking facts, is a huge challenge to professionals in particular. This therefore is shifting knowledge work in the direction of oversight and there is the potential for it to result in deskilling and over-reliance on technology, as well as concerns around intellectual property, data protection, transparency and accountability. These are particularly problematic in high-stakes fields, such as law, healthcare, financial services, journalism and public administration, where even the smallest error of a single fact can have serious consequences. The impact of Generative AI at the organizational level will also depend on several factors, including the work environment, how the AI is implemented by management, the organizational digital environment, and the regulatory environment. All workers will feel the effects of Generative AI differently based on their tasks, level of autonomy and level of skill to apply a Generative AI to complete a task successfully. In some cases, workers may be empowered with tools that lead to increased productivity by helping to redesign work to take greater advantage of the worker. Generative AI could also be used for surveillance, to cut costs, or to replace workers altogether. To fully grasp the impact of Generative AI on knowledge work, it is important to examine how the technology is transforming daily tasks of workers at various levels of skill to complete work. This, in turn, will depend on the degree to which work has been safeguarded against errors, and on the extent of training provided to effectively apply a Generative AI to complete tasks and reach goals at work. This literature review aims to address the question of how generative AI is changing knowledge work. To answer this question the effects of generative AI on the productivity of knowledge work, on the required skills, on the organization in which generative AI is being used and the risks for workers and for organizations associated with generative AI are examined in a comprehensive review of the literature. In this review the changes generated by generative AI for knowledge work are placed in the broader context of automation of work and of professional work.

Generative artificial intelligence (AI) is transforming knowledge work — the creation and application of information — faster than any other wave of automation. It has generated significant interest across a variety of disciplines including management, labor economics, education, organizational studies, and information systems. Generative artificial intelligence is transforming knowledge work. The main focus in this body of literature is generative AI's ability to increase productivity. There is evidence from early studies and field tests that AI tools will significantly decrease the time needed to write, develop code, do research, provide customer support, and prepare decisions for people doing moderately complex jobs. For example, generative AI can help generate drafts of documents; synthesize information to create reports; generate creative ideas; debug software; and translate technical material for non-technical people. In theory, these capabilities will enable knowledge workers to spend more of their time using their judgement; being creative; communicating with other people; and making long-term business decisions. Additionally, there is evidence that AI will close performance gaps by providing inexperienced workers access to guidance, examples and pre-formatted responses that they would normally take years to learn through experience and/or training. There are however a number of limitations and challenges associated with generative AI. For example, generative AI produces output that can be factually incorrect or include bias. Often the biases included in the output are well written and thus hard to identify as false. Thus, users must now verify output as true; critically evaluate output; and ensure that the user understands enough about the subject matter to understand what is correct and what is incorrect. Further, rather than replace the need for humans to make judgments regarding whether a particular action is correct or not, generative AI shifts knowledge worker responsibilities toward ensuring that the actions taken by an organization are ethically appropriate; verifying quality; designing prompts; assessing the appropriateness of actions based upon the context in which they were created; and assessing whether the actions are consistent with the goals of the organization. In addition to the limitations and challenges identified above, concerns exist regarding deskilling, over-reliance on technology, protecting intellectual property rights; protecting individual data privacy rights; identifying who should be held accountable when output produced by an AI system results in harm or injury to a person or group of persons; and lack of transparency surrounding the development process used to create an AI system. The concern regarding deskilling exists because once a person uses a tool for a period of time, the person begins to lose proficiency in performing tasks without the aid of the tool. Over-reliance on technology exists because if a person does not know how to perform a task manually then he/she cannot troubleshoot problems associated with the technology. Protecting intellectual property rights exists because AI systems can reproduce copyrighted materials at very low costs. Data protection laws protect consumers' personal identifiable information. Accountability exists because no one knows how to fix things that go wrong when caused by AI systems. Lack of transparency surrounds the development process of an AI system because many organizations keep proprietary secrets surrounding the methodology used to train their AI systems. Many of these issues are extremely relevant in high-risk areas such as law, medicine, banking and insurance, journalism and government. The organizational literature also points out that the impact of generative AI will depend on how it is deployed. The deployment is influenced by workplace culture; manager practices; digital infrastructure; regulations governing the industry/organization; and workers' skill sets. Therefore, while some organizations are likely to deploy AI to improve employee empowerment and facilitate workflow redesign, others are likely to utilize it primarily for monitoring purposes or reducing costs. Consequently, generative AI is unlikely to have uniformly positive impacts across all occupations or industries. It is expected that the outcomes of implementing generative AI will differ depending on job composition; level of job autonomy; availability of safeguards at the organizational/institutional levels; and worker exposure to formal training programs.

Generative artificial intelligence (AI) - including large language models, multimodal AI, and natural language processing - will greatly alter the way we think about and conduct knowledge work - the work in which you create, analyse, communicate, and apply information (knowledge). In contrast with previous iterations of automation which focused on physical or administrative tasks that were repetitive, AI now automates the creation of text documents, computer code, graphical content, summaries, analysis, recommendations etc. This opens up automation to all forms of work traditionally performed by professionals. Interest in generative AI has thus been expressed in numerous scholarly disciplines, including information systems, management science, labour economics, education, and organisational behaviour. The primary focus of the literature review is to understand how generative AI has the potential to increase productivity or the speed or efficiency with which work is done. As shown by early research studies and field experiments, there is strong evidence that AI technology can reduce the time it takes to create written documents, generate computer programs, conduct research, provide customer service, or plan decision-making. [Consider replacing the following with something like: ". AI will enable knowledge workers to perform tasks such as writing documents, synthesising information from multiple sources, generating new ideas that support the user's objectives, troubleshooting problems in a computer program, and translating specialised content targeted at specified audiences into plain English. AI's capabilities will allow users to devote more time to making informed judgements, exhibiting creativity, building relationships with co-workers and clients, and making strategic organisational decisions. It can also help to level the performance field for less physically capable workers by providing access to the necessary tools (e.g., guidance, examples from which to work, structured output) that would otherwise take a long period of time to learn or would require assistance from an expert.] Conversely, the literature also identifies some of the limitations and risks associated with generative AI. For example, generative AI systems produce inaccurate, biased or fictitious responses, which may be very difficult to identify due to the high degree of fluidity in their language. Subsequently, there will be new requirements for users to verify the information provided by the generative AI system, intended user to critically assess the responses from a generative AI system, and for users to possess the requisite domain knowledge to accurately utilise the content generated by a generative AI system. That said, generative AI may not reduce the reliance on human judgement as much as change the form in which knowledge workers will be required to exercise their judgment (oversight of generative AI outputs, developing prompts for generative AI, quality control of generative AI outputs, and performing ethical evaluations of generative AI outputs). In addition, concerns related to deskilling of the workforce represented by over-reliance on generative AI technology, ramifications of using generative AI technology with respect to intellectual property, user data privacy, accountability for the use of AI, and lack of transparency related to generative AI outputs will need to be considered before adopting generative AI technology in any high-stakes profession (e.g., law, medicine, accounting, journalism, or public service) due to the significant societal and/or professional impact that will result from errors. In addition, research in organisational behaviour suggests that the impact of generative AI will be determined by the manner in which organisations implement it. For example, an organisation's norms and expectations of employees in terms of the acceptance of and ability to use AI, as well as regulatory restrictions related to workplace productivity will influence the extent to which AI is implemented and the degree to which it is actively used by employees. Some organisations will implement generative AI technology for the purpose of improving employee productivity and redesigning the way that their employees work, while others will use generative AI technology primarily as a tool for improving performance measurement, reducing costs, or replacing employees. As such, the impact of generative AI is unlikely to affect all occupations or industries uniformly and the manner in which an organisation will use generative AI technology will depend on multiple factors related to the task, professionalism of the user and the organisation, internal organisational safeguards, and how to stage the introduction of generative AI to the workforce. The literature review synthesises empirical research examining the ways in which generative AI has changed the nature of knowledge work, while situating the generative AI phenomenon within a broader debate relating to automation and professional labour. Consequently, the literature review will provide insight into the opportunities and tensions associated with generative AI.

The introduction of generative artificial intelligence (AI)—specifically large language and multimodal models—into the practice of knowledge work, which entails the generation, understanding, communication, and application of knowledge, has been swift. Unlike previous technological revolutions in automation that mainly impacted routine manual jobs and back-office functions, generative AI can generate text, code, images, summaries, analyses, and recommendations, thereby automating tasks in professional, higher-skill fields. This has led to considerable research attention in information systems, management, labor economics, education, and organization research. There is a significant amount of literature that focuses on the opportunities that generative AI presents to knowledge workers. Research and field experiments find some evidence that generative AI tools can save time writing, coding, researching, customer support, and decision support, for example, in particular settings and for particular levels of task complexity, but more generally for moderately complex knowledge tasks. Generative AI could aid in composing documents, understanding information, brainstorming, troubleshooting code, and translating information to different audiences. This could lead to changes that could potentially enable professionals to devote more of their time and expertise to judgment and critical thinking, creative expression, persuasion, and decision-making. The technology could also potentially help bridge the gap in the performance of novice and expert workers by providing the support and guidance (e.g., in the form of examples and templates) that less experienced individuals might lack or require expert help to acquire. However, there are also numerous examples in the literature that demonstrate the limitations and risks associated with generative AI. The technology could generate factual errors or false or biased information, for example, or provide misleadingly convincing results with poor underlying reasoning, potentially with few outward signs of error. This would lead to new requirements for the verification and scrutiny of information. Moreover, knowledge work is likely not to be entirely automated, but rather may become a task to provide oversight (i.e., to check the quality of AI-generated material), create prompts, ensure quality, and determine ethical issues, with potentially more of a need for domain expertise. There could be risks in deskilling, dependence on the technology, intellectual property concerns, data privacy issues, responsibility of output, and the transparency of generative AI. Such issues could be particularly impactful in law, health care, financial management, journalism, education, and government, where erroneous information could have serious social and professional repercussions. Generative AI may also have different organizational implications for knowledge workers. For example, it could be the case that generative AI is used to support and empower workers in the workplace. In this regard, there are a number of organizational and managerial factors and contextual settings, including workplace norms and management approaches, information infrastructure and regulations, skills of workers, that affect the nature of change with generative AI. Thus, it is plausible that the technology has heterogeneous effects across sectors and occupations; generative AI is unlikely to be universally transformative, given that the nature and extent of change depends on the task environment, the degree of knowledge-worker control over workflow, the institutional context, and the training of the worker. This review paper contributes to the understanding of how the introduction of generative AI to the practice of knowledge work by reviewing and synthesizing studies of the generative AI technology, including the literature on productivity, changes in skills and knowledge work, organizational adoption, and ethical issues. The review paper provides an understanding of generative AI to knowledge work, situating it within the broader discussion of automation and professional work and helping to identify both the opportunities and challenges of this novel technology.

Generative AI – large language models and multimodal systems – has quickly transformed knowledge work, namely work involving production, processing, communicating and applying information. In contrast to previous automation that has mostly impacted routine physical and clerical jobs, generative AI produces texts, programs, images, summaries, analytics and recommendations thus automating tasks previously associated with professional expertise. That is why the appearance of generative AI has triggered much discussion in information systems, management, labor economics, education, and organizational studies. One of the most common topics in the literature is productivity potential of generative AI. The early studies and field experiments show that using AI tools can decrease time needed for writing, programming, research, customer support, and decision preparation especially when workers perform complex enough tasks. Generative AI can help in producing documents, synthesizing information, generating ideas, debugging code and translating technical material into understandable language for other people. Such possibilities make it possible for knowledge workers to pay more attention to judgment, creativity, communication and strategy-making. There is some evidence that AI support could help close the performance gap and provide less experienced workers with necessary guidance and examples of outputs that usually require long-term training. The limitations of generative AI technologies discussed in the literature include its tendency to generate inaccurate, biased or even false information that is fluent linguistically making it hard to distinguish. It means that in addition to producing information, the generative AI technologies increase the need for verifying it and critical evaluation in general. Instead of making human judgment unnecessary, generative AI shifts knowledge work towards supervision, prompt generation, quality control and ethical assessment. The problems with deskilling, overreliance, intellectual property, data privacy, accountability and opacity of AI-generated outputs are widely discussed. All those problems become even more relevant in high-stake industries like law, medicine, finance, media and public administration because of social and professional consequences of the mistakes. According to organizational literature, generative AI implementation depends on workplace culture, managerial practices, existing digital infrastructure, regulation and workers' skills. It is possible that some organizations use AI technologies to empower workers and reshape workflows while others use it just to monitor their activity or cut costs. It is unlikely that generative AI would have uniform consequences across occupations and industries. Task composition, autonomy of professionals, institutional protections and learning opportunities matter in this case. This literature review looks at transformation of knowledge work in relation to generative AI considering research on productivity, skill change, organizational adoption and ethical risks.

Generative artificial intelligence (AI) is transforming the landscape of knowledge work—defined as tasks involving the creation, interpretation, communication, and application of knowledge—in a fast-paced manner, especially with the advent of large language models and multimodal systems. AI previously automated mostly physical or administrative repetitive tasks; however, with the advent of generative AI, advantages can be realized for tasks that would have been in the realm of professional expertise such as writing text, coding, generating images, and creating summaries, analyses, and recommendations. It is therefore both its emergence and its significance that has attracted considerable interest in such diverse disciplines as information systems, management, labor economics, education, and organizational studies. One of the common issues present in the literature is the ability of generative AI to help with productivity. Preliminary research and field tests indicate that AI capabilities can cut down on time spent writing, coding, conducting research, providing customer service and preparing decisions, particularly among employees that carry out somewhat complex tasks. Generative AI can help you draft documents, synthesize information, come up with ideas, debug software and translate technical content for various audiences. Such skills could enable an increased concentration of judgment, creativity, inter-personal communication, and strategic decision making by the knowledge worker. Additionally, the use of AI for deep reading allows for the production of more concise reports that deliver key insights and content without lengthy manual analysis.What's more, AI's support can help close the performance gap by enabling less experienced team members to deliver more detailed reports, with guidance, examples and structured outputs that might have required intensive manual training. But the literature also presents crucial limitations and risks. Generative AI systems have the potential to generate inaccurate, biased, or fabricated information, which can be persuasive and hard to discern, particularly because they can have high linguistic fluency. This enforces new requirements of verification, critical evaluation, and expertise of the domain. Generative AI could reduce the time people spend on knowledge work, transforming it into tasks like oversight, prompts, quality control, and ethical evaluation. Issues also have emerged relating to deskilling, overreliance, intellectual property, data privacy, and even accountability and opacity of AI-generated outputs. Such problems are paramount in high-risk industries like the practice of law, health care, banking, journalism, and government service, where mistakes may have serious social and professional implications. Additionally, organizational research indicates that the adoption of generative AI will vary based on its implementation. A workplace's norm, managerial strategies, digital infrastructure, regulatory environments and workers' skills influence the adoption of something. Some businesses might turn to AI to elevate business personnel and change processes; others might simply utilize it to monitor things, reduce costs and even replace workers. Thus, the impact of generative AI is unlikely to be consistent across occupations or industries. The impact will depend on how tasks are composed, the level of professional autonomy, institutional protections, and training opportunities. This literature review explores the impact of generative AI on knowledge work by synthesizing the findings of related research on productivity, the need for skill transformation, organizational adoption, and moral risks. The review attempts to provide a sense of the opportunities and tensions through positioning generative AI in the larger context of professional labour and automation.

Generative artificial intelligence (AI)—large language models and multimodal systems—automates aspects of knowledge work. Knowledge work involves accessing, understanding, producing, and using information. Previous technological automation has tended to focus on manual or rote cognitive tasks. In contrast, generative AI can generate text, code, images, summaries, analysis, and suggestions. Because knowledge work has been viewed as immune to automation or reserved for highly educated professionals, generative AI has drawn interest from information systems, management, labor economics, education, and organization scholars. Research has centered on topics such as productivity impacts, how generative AI may change skills requirements for knowledge work, how organizations are adopting generative AI, and ethical concerns surrounding its use. This literature review provides an overview of research on how generative AI is transforming knowledge work.

Generative artificial intelligence (AI), the next stage in AI technology characterized by large language models and multimodal systems, is transforming knowledge work — tasks that pertain to creating, interpreting, communicating and applying information. Generative AI represents a new paradigm of kind of work - unlike previous waves of automation that broadly harmed non-cognitive and clerical jobs, generative AI creates text, code, images/summaries/analysis/recommendations/open context– placing automation in domains that have historically been reserved for professionals. Consequently, it has stirred significant interests in information systems, management, labor economics, education and organizational studies. The potential of generative AI to increase productivity is a dominant theme in the literature. Existing evidence based on studies and field experiments indicates that the time needed to write, code, research, provide customer support or prepare a decision can be reduced by AI tools at least for workers performing moderately complex tasks. It can help draft documents, synthesize information, generate ideas, debug software and translate technical content for varied audiences. This perhaps enables knowledge workers to devote proportionately more attention to judgement, creativity, folksy human interaction and the larger strategic decisions. Research shows that AI can open up guidance, examples and structured outputs to workers who might otherwise need extensive training or the help of an expert, potentially closing performance gaps. The literature also suggests substantial limitations and risks, though. Generative AI systems can generate false or biased information, reports of non-existent churches (with seemingly fluent text), etc — but with text which is so obstensively fluent it becomes difficult to identify those errors. It creates fresh requirements for verification, scrutiny and know-how within. Instead of removing judgment from the human equation, generative AI can lead knowledge work into roles more focused on oversight, prompt creation, quality assurance and suitability. AI systems have also attracted concerns over ceaseless deskilling, algorithmic overreliance, intellectual property valuation debates, data privacy compliance issues, the uncertainty of accountability and the opaqueness of any consequent AI-generated outputs. In high-risk domains like law, healthcare, finance and public administration where errors can lead to grave social and professional ramifications. Organizational research also implies that the effects of generative AI will vary with its implementation. Workplace Norms, Managerial Strategy, Digital Infrastructure, Regulatory Environment and Workers’ Skills all shape adoption. Although some organizations apply AI to enhance skills and reconfigure workflows, others deploy it largely for continuous oversight, expenditure reduction, or replacing human labor. Consequently, generative AI will most likely not impact every occupation or industry in the same manner. Its significance will depend on the assembly of tasks, expert independence, institutional protections, and access to training. This literature review synthesizes research on productivity, skill change, organizational adoption, and ethical risk to inform our understanding of how generative AI is changing knowledge work. The review seeks to elucidate the potential and tensions of generative AI, firmly locating it within wider debates of automation and professional work.

Generative artificial intelligence (AI) — especially large language models and multimodal systems, like ChatGPT that can accept both text and image inputs — is quickly transforming knowledge work: the use of information to create something new or meaningfully process it. While previous waves of automation disrupted largely routine physical or administrative work, generative AI introduces a new era by being able to generate text, code, images as well as summaries and analyses and recommendations across fields that have historically been deemed highly knowledge-intensive professions. This has received almost widespread attention through information systems, management, work economics labor education and organizational studies. A common thread throughout the literature is generative AI’s ability to enhance productivity potential. Forthcoming research and field trials find evidence that AI tools can shorten the time spent writing, coding, researching, clarifying customer questions or preparing basic decision making—if you are a worker of moderate complexity level. Generative AI can help you draft documents, synthesize information or ideas, debug software and translate technical content for lay audiences. Thus these capabilities—and others we will see—may help knowledge workers spend more time on judgment, creativity, interpersonal communication and strategic decision-making. Another stream of research shows that, when less experienced workers receive AI assistance, they are able to obtain guidance and examples or produce outputs in a highly structured way; this helps them close performance gaps—something that could otherwise require extensive training from professionals. The literature also notes important limitations and risks. Generative AI systems can generate text that is wrong, biased or factually incorrect — often at a high level of linguistic fluency making it challenging to identify errors. It also raises new challenges in terms of verification, critical assessment and subject-encompassing knowledge. Instead of rendering human judgement unnecessary, generative AI might reorient knowledge work towards supervision and prompting design along with quality checks and ethical assessments. So, there are concerns concerning deskilling, overreliance, intellectual property and data privacy issues; accountability for the actions (remedying damages caused by AI) as well as opacity of output from these systems. Such issues are especially critical in areas such as law, medicine and healthcare, finance, journalism or public administration —that all have serious social & professional consequences if the algorithm is wrong. Research in organizational studies, however, has found that the effects of generative AI will largely depend on how it is implemented. It is determined not merely by a companys capital or eagerness of its individual workers but also at this level as to how workplace norms, managerial strategies and digital infrastructure interact with the regulatory environment and workers skills. Certain organizations will use AI to empower workers and redesign work; others, however, might deploy it simply for surveillance (e.g., reading employees' emails), or for cost savings/replacement labor. This means that generative AI will not likely affect all occupations or industries evenly. The effects will depend on task composition, professional autonomy and institutional safeguards in the workplace, as well as access to training. The paper, which situates generative AI as a productive force in knowledge work through reviewing literature related to productivity impacts and skill change alongside organizational adoption strategies including provided guidelines for ethical risk mitigation. The review seeks to better define both the opportunities and tensions of generative AI, by placing it in broader conversations about automation as well as professional labor.

Generative artificial intelligence (AI) - specifically large language and multimodal models - is fundamentally altering how much of knowledge work, defined as work involving the creation, interpretation, communication, and application of information, is done. This differs from previous waves of automation, which primarily affected routine physical and administrative tasks; generative AI can now produce text, code, images, summaries, analyses, and recommendations, extending automation into realms previously defined by expert skills and professional discretion. This has led to enormous interest in generative AI among information systems, management, labor economics, education, and organizational scholars. A prominent theme in the emerging literature is generative AI’s potential to boost worker productivity. Early work and field experiments show AI can reduce time costs on a wide range of tasks including writing, coding, research, customer support, and preparing decisions - particularly for workers on moderately complex tasks. For instance, generative AI can draft documents, synthesize text, brainstorm ideas, debug software, and explain technical concepts to nontechnical audiences, thereby freeing up time for more judgment- and creativity-based or interactive elements of work. Moreover, there is evidence that AI tools might level the playing field by helping workers who have less experience in the tasks with guidance, examples, and structured outputs. However, the literature also emphasizes serious limitations and risks. Because AI can generate plausible, fluent output - but it cannot "know" and thus is prone to produce errors or fabrication with confidence, much knowledge work will shift to verification, evaluation, and judgment. We will likely shift from creating the knowledge, to managing AI systems to help create knowledge and then verifying and applying the generated knowledge, requiring higher levels of critical thinking, human supervision, prompt-craft, and accountability. Serious concerns also arise regarding deskilling, overreliance, intellectual property, data privacy, and accountability - especially given AI's opacity in key professional fields like law, medicine, finance, journalism, and public administration. The impact of AI on the organization of work is likely contingent on implementation: workplace norms, manager strategies, legal frameworks, the organization's technological capabilities, and the skills of workers will all shape whether generative AI will be used to augment employees and redesign workflows or replace them, control them, and monitor their work. Hence, generative AI is unlikely to produce uniform effects on knowledge work, varying significantly across occupations and industries depending on work structure, autonomy and institutional design. This paper reviews and synthesize this emerging literature on how generative AI is affecting knowledge work by examining the issues of productivity, skills and labor markets, organizations and work, and ethical risks. It also places generative AI within broader debates on automation and professional labor, both to highlight its promises and its tensions.

Generative AI—especially large language models and systems that handle text and images—has started to transform knowledge work in a big way. This kind of work isn’t just basic office tasks; it’s about creating, analyzing, and sharing information. What’s different this time is that AI isn’t just automating repetitive jobs. It’s writing text, generating code, creating images, summarizing long reports, and even offering recommendations—stuff that used to require a human expert. Naturally, people in fields like management, labor economics, education, and organizational studies are paying close attention. A lot of the research so far talks about how generative AI can boost productivity. Early studies show these tools help people write, code, research, handle customer support, and prep for decisions a lot faster—especially when tasks aren’t super simple but not crazy complex, either. You can ask AI to draft documents, sum up information, come up with ideas, debug code, or explain complicated concepts in plain language. All this gives knowledge workers more room to focus on making good calls, thinking creatively, communicating with others, or planning next steps. Some researchers even say AI help levels the playing field, giving less experienced folks access to the kind of support or examples they’d otherwise have to get from seasoned experts. But, of course, there are drawbacks—and the literature doesn’t shy away from them. Generative AI makes mistakes. Sometimes, it spits out information that’s wrong, biased, or straight-up made up. The worst part: it usually sounds convincing, so it’s easy to miss the errors. That means people need to double-check its work—and maybe spend more time evaluating, fact-checking, or bringing their own expertise. So, instead of taking judgment out of the loop, AI shifts the job toward oversight, prompt-writing, quality control, and ethics. There’s also real worry about people losing skills, relying too much on the technology, copyright headaches, privacy, accountability, and not being able to understand how the AI comes up with its answers. These problems matter a lot in sensitive areas like law, healthcare, money, journalism, and government—anywhere mistakes can come with a high price. Organizational research points out that AI’s impact really depends on how companies use it. Culture, management style, tech setup, rules, and workers’ skills all shape what happens when AI rolls out. Some places might use AI to help workers grow and rethink workflows. Others might see it as a way to monitor staff, cut costs, or even replace people. So, the effects won’t be the same everywhere. They’ll depend on what tasks people do, how much freedom they have, what protections are in place, and whether staff get proper training. This review digs into how generative AI is changing knowledge work by rounding up research on productivity, skills, workplace adoption, and ethical risks. By connecting these dots to wider debates about automation and professional labor, the aim is to get a clearer picture of both the promise and the problems that come with generative AI.

Generative artificial intelligence (AI), especially large language models and multimodal systems, is quickly changing knowledge work. This type of work involves creating, interpreting, communicating, and applying information. Unlike previous rounds of automation that mainly impacted routine physical or administrative tasks, generative AI can produce text, code, images, summaries, analyses, and recommendations. It brings automation into fields that usually require professional skills. Its rise has sparked significant interest in various areas, including information systems, management, labor economics, education, and organizational studies. A key focus in the research is the potential of generative AI to boost productivity. Early studies and field experiments show that AI tools can cut down the time spent on writing, coding, research, customer support, and decision preparation, particularly for workers handling moderately complex tasks. Generative AI can help draft documents, synthesize information, generate ideas, debug software, and translate technical content for different audiences. These abilities may let knowledge workers concentrate more on decision-making, creativity, communication, and strategic thinking. Some studies also suggest that AI can help less experienced workers perform better by providing guidance, examples, and structured outputs that would usually need extensive training or expert help. However, the research also points out significant limitations and risks. Generative AI systems can generate inaccurate, biased, or made-up information, often sounding fluent enough to make mistakes hard to catch. This raises the need for verification, critical evaluation, and domain expertise. Instead of removing the need for human judgment, generative AI may shift knowledge work towards overseeing, designing prompts, ensuring quality, and assessing ethical concerns. There are also worries about deskilling, overdependence, intellectual property, data privacy, accountability, and the lack of transparency in AI-generated outputs. These issues are especially important in high-stakes areas like law, healthcare, finance, journalism, and public administration, where mistakes can have serious consequences. Research in organizations suggests that the impact of generative AI will depend on how it is put into practice. Adoption is affected by workplace norms, management strategies, digital infrastructure, regulatory conditions, and workers’ skills. While some organizations may use AI to empower their employees and rethink workflows, others may use it mainly for monitoring, cutting costs, or replacing labor. As a result, generative AI is unlikely to affect all occupations or industries in the same way. Its effects will differ based on task composition, professional autonomy, institutional protections, and access to training. This literature review looks at how generative AI is changing knowledge work by summarizing research on productivity, skill development, organizational use, and ethical risks. By placing generative AI within larger discussions about automation and professional labor, the review seeks to clarify its possibilities and challenges.