Lit Review · Claude Sonnet 5

Generative AI is one of the biggest changes to knowledge work in many decades. Generative AI can do work like writing code, research summaries, emails, etc. Generative AI can also write code, which means we now have machines that may be able to write code better than us, or at the very least as good as us (and probably better). There is an active field of literature examining the effects of generative AI on knowledge workers. Some key studies in the area include Brynjolfsson, Li, and Raymond (2023), Noy and Zhang (2023), which show that generative AI significantly increases the productivity of knowledge workers, perhaps in an unprecedently broad scope, which may lead to the reduction of productivity dispersion between workers.

Automated software tools have historically made aspects of knowledge work easier or eliminated entire jobs. However, the recent arrival of generative AI changes everything; for the first time we have machines that can meaningfully participate in nonroutine tasks like writing essays, summarizing research, drafting emails, writing code, etc. This has huge implications for the future of knowledge work, and researchers across multiple disciplines (economics, IO, info systems, management, etc.) have begun to study these tools and their potential effects. Here, I provide a literature review that summarizes what we know so far. Knowledge of this literature is constantly evolving at the same pace as these AI tools themselves, so this review will focus on the higher level questions being asked and mark points where literature is conflicting. Initial research has focused on productivity of knowledge workers with AI tools. Brynjolfsson, Li, & Raymond (2023) and Noy & Zhang (2023) both found substantial increases in knowledge worker productivity using AI tools. Interestingly, this effect was larger for lower performing employees, which has implications for within-firm productivity differences. Following this research, there are a number of questions that researchers are looking at. How do AI tools impact the quality vs quantity of knowledge work? How do AI tools impact human capital accumulation? Do these tools change knowledge workers' epistemic agency? In answering these questions, there is some disagreement in the literature. One major question is whether AI tools are a complement to workers or a substitute. The answer has major implications for whether these tools will augment jobs or displace them. Another question is whether AI tools are actually improving the quality of work or simply increasing the speed of work. A third question that is only just being asked is what the secondary effects of these AI tools will be. Will they atrophy certain human skills? Who actually benefits from these productivity increases, workers or owners?

# The Impact of Generative AI on Knowledge Work: A Literature Review Generative AI has caused rapid and drastic changes to most aspects of knowledge work in ways that haven't been seen since knowledge work was fundamentally changed by the introduction of personal computers. Generative AI consists of large language models, or LLMs, like GPT-4 and Claude, among others. Compared with previous automation technologies, those focused on removing the need for repetitive and codified tasks, Generative AI disrupts systems similar to cognitive thought and creativity that humanity has previously monopolized. Systems like these are capable of drafting student essays, writing and debugging code, research, and even artistic endeavors. Because of this extreme disruption of knowledge work, many researchers in different fields have had to rethink original theories regarding the future of work, such as economics, labor studies, and organizational behavior. The pace of research in this field is out matching many other fields and even technological advancement. The first research focused mainly on productivity. In his/her/their works, Brynjolfsson, Li, and Raymond (2023) and Noy and Zhang (2023) focused on generative AI and noted its tendency to improve the performance of knowledge workers, especially for the lower performing workers, to the point of suggesting the existence of a “leveling effect” to AI, whereby the organizational productivity gap may be closed. Questions of augmentation, of the balance between qualitative and quantitative outputs of knowledge work, are all relevant in the context of the development of generative AI. The same goes for the effects on the development of cognitive skills and Human Capital, if generative AI becomes the dominant partner in the performance of cognitive tasks. The most important questions, however, are the ones pertaining to the changes in self-identity of knowledge workers and the professional domain of knowledge work. This research demonstrates several areas of incongruence, the first and perhaps most important being whether generative AI augments or substitutes human capital, an argument which would impact fields and professions as disparate as construction and creative fields. The second would be whether the new AI increases productivity and if that productivity is even measurable in qualitative changes in the work, or if generative AI accelerated the routines of work without real qualitative improvements. Finally, the effects of generative AI may signal the end of critical thought, the beginnings of a uniform style of work, and an imbalance in the professional domain of cognitive work. This review integrates multiple fields of study to evaluate the impact of generative AI on knowledge work, highlighting where scholars currently agree and where research is lacking and is needed.

# The Impact of Generative AI on Knowledge Work: A Literature Review The use of generative artificial intelligence (AI) in knowledge work will be one of the biggest changes to work with computers since the personal computer. Previous automation has, by and large, focused on codifiable tasks that can be performed on a routine basis. In contrast, a new generation of systems – large language models (LLMs) – are designed to handle a wide array of non-routine, non-standardized, cognitive tasks that are typically the domain of a human expert. And so, there is considerable interest in the effects of generative AI from an economic, organizational, information system and labor perspective. Many disciplines are represented by researchers studying these effects. In addition to the issues of how generative AI affects the kinds of tasks that knowledge workers do on a daily basis, and how human workers interact with generative AI systems, there are many recent papers that study a variety of other dimensions of the use of generative AI for knowledge work, such as: (1) the quality vs. quantity of work produced by humans and by humans augmented by generative AI; (2) the skills and ability requirements for work that is supported by generative AI; and (3) the ways in which generative AI is likely to affect the split of value between labor and capital in industries in which knowledge work is a very large part of the total value added by the firm. These papers, in turn, probe in considerable detail several of the key tensions apparent in the prior work on generative AI for knowledge work, for example: (1) whether generative AI as work aid vs. generative AI as work substitute; (2) work with generative AI of higher quality vs. work with generative AI merely of greater quantity; and (3) how generative AI will change the nature of the work performed by knowledge workers on a daily basis. The literature on the impact of generative AI on knowledge work is evolving rapidly, and it creates a number of tensions. First, the literature is divided on the extent to which generative AI will complement human expertise or substitute for it. If generative AI mainly substitutes for human work, then it could increase productivity, but it also risk displacing workers. There is already a substantial body of literature that explores the impact of generative AI on a wide array of professions, including doctors, lawyers, financial analysts, and managers. A second set of issues relates to the relationship between increased productivity and the quality of work. As workers using generative AI become much more productive than they were before, there is the potential for them to be producing work of lower value. In order to understand this potential, there is now a growing body of research that is exploring the so-called “second-order effects” of generative AI on professionals and their work. These second-order effects could include a wide array of negative consequences, including the loss of critical thinking skills by professionals using generative AI. They could also include the production of work that is far too similar in order to avoid errors that could have negative consequences. Finally, generative AI has the potential to dramatically alter the distribution of value between labor and capital in industries that are heavily dependent on knowledge work. In other words, professionals using generative AI could capture much less of the value that they create than they did before. This review aims to synthesize the current body of research that studies the impact of generative AI on knowledge work, and thereby provides a comprehensive overview of the current state of research and identifies promising fields for future research.

# The Effects of Generative AI on Knowledge Work: A Review of Literature The emergence of generative artificial intelligence has triggered one of the largest paradigm shifts in the field of knowledge work ever since the invention of the PC. While the earlier waves of automatization mainly eliminated routine and codifiable tasks, generative AI systems such as GPT-4, Claude, and other similar technologies have shown impressive performance in activities that have been deemed exclusively human before: writing sophisticated documents, generating codes, conducting research, and even producing creative content. This newfound ability to automatize or augment non-routine cognitive tasks has attracted considerable attention from the scientific community representing economics, organization behavior, information systems, and labor studies and forced researchers to revise their assumptions concerning the future of professional work. The academic discussion of this phenomenon has developed at the same breakneck speed as technology itself. The first studies were focused on the implications of generative AI on the productivity of knowledge workers, and the landmark works by Brynjolfsson et al. (2023) and Noy & Zhang (2023), who discovered impressive performance improvement on the part of professionals working with these technologies, mainly lower-performers, suggested a possible "leveling effect" of AI-powered software, potentially eliminating disparities in performance within organizations. More recent research has started addressing a number of additional, but important questions, such as: what is the effect of augmentation by means of generative AI on the quality of output produced by knowledge workers? What kind of consequences could the automatization of routine cognitive tasks have on skill formation and human capital? How does knowledge workers' professional life change with the rise of AI-powered technology? It seems that several conflicting trends can be discerned in the academic discussion of this issue. First, there are disputes in the scientific community as to whether generative AI augments human capabilities or acts as a substitute for it, leading to different conclusions regarding its impact on job augmentation/displacement within a number of professional fields. Second, there is an ongoing debate regarding the implications of productivity improvements on the qualitative changes in the knowledge work, that is, if the processes are accelerated by the new technology but are not improved in their essence. Third, more recent research addresses the issue of second-order effects related to the use of generative AI: loss of skills of critical thinking, homogeneity of professional output, and redistribution of value between labor and capital within knowledge-intense industries. This paper analyzes this body of literature with the aim to conduct a thorough assessment of generative AI's impact on knowledge work and identify both its consensus and problems that need further elaboration.

A literature review of the effects of generative AI on knowledge work. With the arrival of generative artificial intelligence, one of the biggest changes in the face of knowledge work since the personal computer has emerged. In stark contrast to the automation that automated routine, codifiable tasks, generative AI systems like large language models (LLMs) such as GPT-4, Claude, and their successor have shown remarkable capabilities, engaging in tasks that have traditionally been the exclusive domain of human cognition — from drafting complex documents and generating code to synthesizing research and even creating creative content. Such unprecedented power to augment and automate non-routine cognitive tasks has created significant interest among many researchers spanning a variety of fields such as economics, organizational behavior, information systems, and labor studies, leading to a re-thinking of basic notions about the future of professional work. There has been a huge amount of literature on this topic in the past few years, following the rate that technology has been advancing. The early research tended to emphasize productivity effects, and we do have some good illustrations of the performance improvements of knowledge workers employing generative AI tools, especially for those that had the lowest performance, which could imply a “levelling effect” and consequently a decline in productivity gaps within organizations, such as in Brynjolfsson, Li, and Raymond (2023) and Noy and Zhang (2023). Follow-up studies have investigated increasingly subtle questions, like what does AI augmentation mean for the quality of knowledge work output as compared to quantity? If there are more routine cognitive tasks delegated to AI systems, what are the implications for skill development and human capital formation? But most importantly, how do these tools, transform the epistemic authority, or professional identity, of knowledge workers? This body of literature disclosing a number of tensions which the careful examinee sees. First, there is controversy over the role of generative AI as an augmentation or replacement of human expertise, which could either augment or fully replace jobs at the wholesale level in different fields of work. Secondly, there are conflicting views in the literature as to whether productivity gains equal significant improvements in the quality of work or simply help speed up work processes without actually increasing gain in the quality of the work product. Third, the secondary consequences, in terms of a potential decline in critical thinking skills, the product's homogenization, and a change in the ratio between labour and capital in knowledge-intensive sectors are issues that emerging research is increasingly tackling. This review aims to integrate these threads of research to provide an overall picture of the impact of generative AI on knowledge work and a sense of areas of consensus and enduring gaps that require more study.

Generative Artificial Intelligence: Implications for Knowledge Work Generative artificial intelligence (AI) is driving one of the most significant changes to knowledge work since personal computers were introduced. Rather than automating primarily repetitive, well-defined tasks, generative AI systems (e.g., large language models [LLMs] like GPT-4, Claude, and others) have displayed impressive abilities to perform non-routine cognitive tasks that have been thought to require human intelligence—creating complex documents, writing computer code, aggregating research from various sources, and even developing creative works. The unprecedented ability of generative AI to either enhance or replace non-routine cognitive work has led to significant interest in the academic community across a range of disciplines (economics, organizational behavior, information systems, labor studies, etc.) as researchers work to rethink long-held beliefs about the future of knowledge work. The literature on this topic has evolved dramatically in line with the rapid pace of innovation. Initial studies have focused on the potential productivity effects of generative AI, with many reports documenting significant increases in the productivity of users of generative AI tools (especially for users who have historically performed well below average)—indicating a "leveling effect" on productivity performance within organizations. As more researchers begin to investigate generative AI's potential throughout a variety of sectors, more nuanced questions have arisen: How will AI enhancement impact the quality of work vs. the quantity of work product? Will human capital and skill formation be impacted by increasing the extent to which routine cognitive tasks are assigned to AI? And perhaps most importantly, how will generative AI change knowledge workers’ professional identities and epistemic authority? The emerging body of literature has identified several areas of tension that require additional investigation. For example, researchers are split on the question of whether generative AI primarily acts as a supplement to human skills or as a replacement; thus, some professionals in the sectors will find their jobs enhanced and others displaced entirely. Additionally, there is disagreement on whether productivity gains have actually improved the overall quality of work or merely increased the speed with which work is completed without materially improving the quality of the work produced. Finally, many researchers are beginning to address the second-order impacts of generative AI use, including the potential loss of critical thinking skills, the homogenization of outputs produced by professionals, and the potential reallocation of capital and labor value within knowledge-centric industries. This literature review will synthesize the existing literature on the implications of generative AI for knowledge work to evaluate areas of consensus as well as gaps that warrant future study.

Generative AI's impact on knowledge work: a literature review The rise of generative AI is triggering one of the biggest changes in how knowledge work gets done since personal computers came along. While previous waves of automation have focused mainly on eliminating routine, codifiable tasks, generative AI systems — including state-of-the-art large language models (LLMs) like GPT-4, Claude and their successors — have shown remarkable potential in areas that were once thought to be uniquely human: drafting complex documents, generating code, synthesizing research results and even producing creative content. This unprecedented ability to enhance or automate non-routine cognitive tasks has sparked enormous academic interest in fields as diverse as economics, organizational behavior, information systems and labor studies, forcing researchers to question basic beliefs about the future of professional work. Evidence on this issue has rapidly evolved, in tandem to the rapidity of technological development itself. Initial research has focused on productivity effects—with notable results highlighting how generative AI tools appear to boost performance particularly among lower-performing knowledge workers by Brynjolfsson, Li and Raymond (2023) as well as Noy and Zhang (2023)—possibly suggesting a “leveling effect” reducing organization-wide productivity disparities. Innovations in AI have led to more recent research exploring less blunt questions: Does AI augmentation influence how good versus how much knowledge work gets produced? What human capital and skill development repercussions will be seen when we delegate cognitive tasks that are routine to AI systems? And crucially, how do these tools reconfigure the epistemic authority and identity of knowledge workers? This body of research uncovers several tensions that need to be carefully examined. One is within the scholarly debate over whether generative AI behaves more as a complement to, or substitution for human intelligence — with potential consequences ranging from partial job augmentation to full-scale displacement in multiple industries. Second, the literature exposes a tension between whether productivity rates yielded tangible increases in work quality or were simply topological accelerants of pre-existing processes without substantive gains. Third, newer research is also beginning to deal with second-order effects like the erosion of critical thinking, dependence on professional written outputs becoming indistinguishable across systems, and how value will shift between labor and capital in knowledge industries. In this review, we synthesize these strands of research to deliver an integrative assessment of the implications for knowledge work with generative AI, highlighting both areas where consensus exists in the literature and enduring gaps that require additional empirical work.

Generative AI represents an epochal shift in the nature of knowledge work — potentially comparable to that created by the advent of the PC. Previous waves of automation have largely replaced routine, codifiable tasks; however, generative AI (large language models [LLM] e.g., GPT-4, Claude, etc.) is capable of performing many previously uniquely human cognitive tasks — creating sophisticated documents, writing code, synthesizing research findings, and even generating creative content. The ability of generative AI to augment or replace non-routine cognitive tasks has attracted considerable attention across multiple academic disciplines — economics, organizational behavior, information systems, and labor studies — and has led researchers to challenge some of the foundational beliefs regarding the future of professional work. The literature on this subject has developed at an extraordinary rate — similar to the rapid development of technology itself. Initial research focused mainly on understanding the productivity benefits associated with generative AI — specifically Brynjolfsson et al. (2023) and Noy & Zhang (2023), who found generative AI had positive impacts on performance for all employees — but especially those with lower levels of productivity. While early research provided evidence of the "leveling effect," where generative AI helped to narrow productivity disparities between employees — subsequent research began to explore more nuanced questions. For example: Does AI increase the quality vs. quantity of knowledge work outputs? How will generative AI influence employee learning and development opportunities when routine cognitive tasks are increasingly being performed by AI systems? Perhaps more importantly, how will generative AI impact the perceived level of authority/knowledge and professional identity of employees engaged in knowledge work? Collectively, these lines of research reveal several important areas that require close consideration. First, there is disagreement over whether generative AI serves as a complement to human expertise or replaces it entirely — which has significant implications depending on the profession/industry involved. Second, there is debate surrounding whether increased productivity translates into improved quality of work outputs or simply accelerates current processes without adding new value. Finally, recent research is beginning to address second-order consequences of generative AI use in knowledge work settings — including the possible erosion of critical thinking skills among employees, the loss of unique characteristics in professional outputs, and the distribution of value between labor and capital in knowledge-based industries. Overall, this paper provides a synthesis of the above research lines to present a comprehensive evaluation of the impact of generative AI on knowledge work — highlighting areas of agreement as well as persistent gaps that should be explored through additional research.

# The Impact of Generative AI on Knowledge Work: A Literature Review The advent of generative artificial intelligence marks one of the most profound disruptions to knowledge work since the advent of the personal computer. Whereas earlier automation efforts largely displaced codifiable, routine tasks, the generative capabilities of artificial intelligence systems-including LLMs such as GPT-4 and Claude-have recently reached an astonishing level of sophistication in areas once believed to be the exclusive preserve of the human mind: writing, coding, synthesis and even creation. The sheer novelty and power of these systems’ capacity to supplement or replace a non-routine cognitive tasks have ignited widespread interest among researchers from economics to information systems, from organizational behavior to labor studies, challenging fundamental presumptions about the nature of professional labor. The research to date on this topic has proliferated at an almost blinding speed, matching the technology itself. Early studies often emphasized productivity impacts, with pioneering work by Brynjolfsson, Li and Raymond (2023) and Noy and Zhang (2023) demonstrating substantial improvements in knowledge work productivity, and showing this effect to be larger among lower-performing workers (a kind of "leveling" effect that could reduce firm-level wage inequality). More recent work has sought to refine the analysis, inquiring: under what circumstances is work quality, rather than work quantity, improved by generative AI? What are the implications of delegating the production of certain routine cognitive tasks to AI for learning, the formation of human capital and the acquisition of skills? How, critically, do these tools reconfigure epistemic authority and professional identity among knowledge workers? Existing literature reveals several competing points of view. One major area of tension concerns whether generative AI is best understood as a complement to human labor or as a substitute; the implications differ for various occupational and industry sectors. Furthermore, although numerous studies indicate substantial increases in output quantity, the extent to which these translate into increases in the quality or substantiveness of work is less clear. And third, the latest work has begun to focus more intently on secondary impacts, including whether generative AI can erode critical thinking, cause the homogenization of professional products and lead to new distributions of the value of knowledge work between capital and labor. This literature review provides a framework for assessing and summarizing these trends, outlining current scholarly agreement, and pointing toward important avenues for future research.

Title: The effect of generative AI on knowledge work applied to the field : A New Lens Literatur Review Generative AI has triggered one of the most consequential paradigm shifts in knowledge work since the introduction of personal computing. In contrast to earlier waves of automation that displaced primarily routine, codifiable tasks, generative AI systems—large language models (LLMs) like GPT-4 and Claude or their successors—are in a domain space where they have done very well: domains long considered the province only of human cognition: drafting paperwork; generating code; synthesizing research—and even creative content. This unique ability to enhance or automate non-routine cognitive work has attracted considerable scholarly attention stemming from many domains: economics, organizational behavior, information systems and labor studies – all of which have forced researchers to reconsider foundational assumptions about the future professional career. This topic has been growing quickly, matching the rapid development of technology itself. The earliest studies mainly investigated the productivity effects, with major contributions by Brynjolfsson et al. (2023) and Noy & Zhang (2023): both demonstrate substantial performance improvements for knowledge workers using generative AI tools, especially among lower-performing employees—which is consistent with a possible leveling effect whereby such tasks may be easy enough to move output along an ability-by-task curve downward within organizations. Newer research has broadened to consider more nuanced questions: What is the effect of AI augmentation on knowledge work output — quality versus quantity? When routine cognitive work is progressively offloaded to AI systems, what does this mean for skill formation and human capital development? How might tools of this kind shift the epistemic authority and professional identity—the strength/characterization of knowledge workers as a realized profession—of those people themselves? This body of research uncovers important tensions to think about. For one, there is disagreement over whether generative AI performs a complementary role to human skill as an augmentor of jobs or more primarily serves as an outright substitute (and all the implications from that point forward ranging across types of roles and entire professional fields). Second, the literature illustrates different points of view on whether productivity gains yield significant enhancements in work quality or merely quicken humans' doing existing processes without adding substantive value. Third, a growing body of research engages with second-order effects stemming from the conventional use and deployment methods employed by generative AIs that ultimately lead to erosion in critical capabilities or homogenization across professional outputs as well as shifts in equity between labor and capital across knowledge-intensive industries. This review synthesizes these strands of research into a holistic evaluation of the impact on knowledge work by generative AI, highlighting where researchers broadly agree and identifying persistent gaps in evidence to guide future empirical investigation.

Generative AI has shaken up knowledge work in ways we haven’t seen since personal computers first hit offices. In the past, automation mostly chipped away at routine tasks—think data entry or basic record-keeping—but generative AI has jumped the fence. Now, these systems, with models like GPT-4 and Claude, can write detailed reports, code up entire projects, condense research papers, and even come up with original creative material. Jobs that once called for deep thinking and creativity aren’t off-limits anymore, and that’s caught the attention of economists, organizational experts, and labor researchers. People are starting to question some long-held ideas about what jobs will look like in the future. Research on this subject has been moving fast, keeping pace with the technology itself. Early on, folks like Brynjolfsson, Li, and Raymond (2023), along with Noy and Zhang (2023), showed that generative AI can make knowledge workers more productive—especially those who usually lag behind. Some call this a “leveling effect,” where gaps in employee performance start to close. More recent studies dig deeper: Does AI just let us churn out more work or does it actually make that work better? If AI takes over the simpler thinking tasks, what happens to human skill-building and expertise? Maybe most important, how does this new tech change what it means to be a professional, or who gets seen as the expert? Of course, there’s no shortage of debate. One big question is whether AI works best as a partner to human experts, or if it’s just flat-out replacing some of them. That’s not just academic—real jobs hang in the balance. Others argue about whether all this new productivity actually leads to better outcomes, or if we’re just doing more of the same without raising the bar. There’s another layer, too: some are worried that as we let AI handle more tasks, critical thinking skills could take a back seat, creative work might all start to look the same, and the way profits and power are divided up could shift—maybe not in favor of workers. This review pulls together these different lines of research to paint a clear picture of how generative AI is changing knowledge work. Plenty of questions still need answers, but we can see where researchers mostly agree and where the big gaps are.

# The Impact of Generative AI on Knowledge Work: A Literature Review The rise of generative artificial intelligence has led to one of the biggest changes in knowledge work since personal computers were introduced. Unlike previous waves of automation that mainly replaced routine, clear tasks, generative AI systems, including large language models (LLMs) like GPT-4 and Claude, have shown impressive abilities in areas long thought to belong only to humans. These include drafting complex documents, generating code, synthesizing research, and creating content. This new ability to improve or automate non-routine cognitive tasks has attracted significant interest from scholars in fields such as economics, organizational behavior, information systems, and labor studies. Researchers are now reevaluating key beliefs about the future of professional work. The literature on this subject has developed quickly, keeping pace with technological advancement. Early studies mainly focused on productivity, with important research by Brynjolfsson, Li, and Raymond (2023) and Noy and Zhang (2023) showing notable performance gains among knowledge workers using generative AI tools. This was especially true for lower-performing employees, suggesting a possible "leveling effect" that could narrow productivity gaps within organizations. Later research looked into more complex questions, such as how AI assistance influences the quality versus quantity of work output. It also examined the effects on skill development and how human capital is formed when routine cognitive tasks are increasingly handed over to AI systems. Moreover, there are important questions about how these tools change the authority and professional identity of knowledge workers. This body of research uncovers several conflicts that need careful analysis. First, there is disagreement among scholars about whether generative AI serves mainly as a complement to human skills or as a substitute, affecting job enhancement or outright replacement in various professional fields. Second, the research shows differing views on whether productivity gains lead to real improvements in work quality or simply speed up existing processes without adding value. Third, emerging studies are starting to address secondary effects, such as the potential decline of critical thinking skills, the uniformity of professional outputs, and changes in how value is shared between labor and capital in knowledge-intensive industries. This review brings together these research threads to give an in-depth look at generative AI's impact on knowledge work. It highlights both areas where scholars agree and gaps that still need further study.