The Future of Work with AI: Moving from Individual Gains to Collective Intelligence

Listen to this article:

Abstract: This report synthesizes recent evidence on how artificial intelligence is reshaping work, drawing from Microsoft's 2025 New Future of Work Report and the broader research literature. While 2024 marked individual productivity gains from generative AI, 2025 signals a critical shift toward collective productivity—how teams, organizations, and communities can improve together with AI. Adoption continues to accelerate globally, with enterprise ChatGPT messages increasing eightfold year-over-year, yet organizational success depends heavily on employee engagement, trust, and participatory design rather than top-down mandates. Evidence reveals meaningful productivity gains and time savings, particularly in knowledge work, but also emerging challenges including AI-generated "workslop," cognitive deskilling risks, and mixed labor market effects concentrated among early-career workers. Human-AI collaboration is evolving from passive tool use to active partnership, requiring new interaction paradigms, robust common ground, and cognitively engaging workflows. Teams face distinct challenges as AI shifts from supporting individuals to enabling group work, demanding new evaluation frameworks, proactive agent behaviors, and careful attention to social dynamics. This report examines adoption patterns, workforce impacts, collaboration design, cognitive implications, and sector-specific transformations while highlighting that AI's ultimate value depends not on technical capabilities alone but on intentional organizational choices that prioritize human agency, skill development, and equitable outcomes.

The fifth annual Microsoft New Future of Work Report arrives at an inflection point. Since 2021, each year has brought transformative change—remote work necessitated by pandemic conditions, the emergence of hybrid arrangements, the introduction of large language models, and their real-world deployment. Yet these are not separate revolutions but chapters in a continuous evolution of how humans collaborate in digital environments.

2024 demonstrated that AI delivers substantial productivity gains at the individual level. Surveyed enterprise users report saving 40–60 minutes daily through AI use, and controlled experiments across writing, coding, customer service, and other domains consistently show time savings and quality improvements (Chatterji et al., 2025; Brynjolfsson et al., 2025). The technology has crossed a capability threshold where meaningful assistance is now routine rather than experimental.

The frontier for 2025 is collective productivity—moving beyond what AI enables for isolated individuals to how it can enhance teams, organizations, and entire ecosystems working together. This shift introduces fundamentally different challenges. Individual assistance requires understanding one person's goals and context; team assistance requires navigating multiple perspectives, managing turn-taking dynamics, building shared understanding, and supporting social processes like trust-building and conflict resolution. Early evidence suggests AI systems trained and evaluated for individual use do not automatically succeed in team settings (Dell'Acqua et al., 2025; Schmutz et al., 2024).

This transition also surfaces important questions about work's social and cognitive dimensions. As AI handles more routine tasks, human contributions increasingly center on judgment, creativity, relationship-building, and the tacit knowledge that comes from lived experience (Autor & Thompson, 2025). Organizations that view AI purely as a cost-reduction tool miss opportunities for augmentation and innovation that could expand possibilities rather than simply automate existing work (Brynjolfsson, 2022).

The stakes extend beyond productivity metrics. How AI gets integrated into work affects skill development, employment patterns, workplace relationships, and whether technological progress broadly benefits workers or concentrates gains among capital owners and high-skill elites. The evidence reviewed here suggests outcomes remain contingent—shaped by organizational choices, policy frameworks, and design decisions that have yet to solidify into dominant patterns.

The AI Adoption and Usage Landscape

Rapid Growth in Enterprise and Consumer Adoption

Generative AI adoption accelerated dramatically in 2024. Enterprise ChatGPT messages increased eightfold over the previous year, while the consumer platform reached over 700 million weekly active users globally by mid-2025 (Chatterji et al., 2025). The gender gap that characterized early adoption has closed—women and men now use consumer AI tools at nearly equal rates, a remarkable shift from early 2023 when over 80% of users were male.

Investment continues to grow alongside usage. Global private investment in generative AI reached $33.9 billion in 2024, an 18.7% increase from 2023, complemented by rising public investment (Maslej et al., 2025). This capital influx reflects widespread conviction that AI represents a fundamental shift in competitive advantage across industries.

Usage patterns reveal which activities see the heaviest AI application. Analysis of ChatGPT conversations found that "Practical Guidance," "Seeking Information," and "Writing" account for approximately 80% of all usage (Chatterji et al., 2025). Research examining Claude usage showed 37% of tasks associated with computer and mathematical occupations (Handa et al., 2025). Analysis of Microsoft Bing Copilot logs revealed that learning, communicating, and writing activities dominate both user goals and AI actions, with knowledge workers in sales, computer occupations, media, and administrative roles showing highest AI applicability (Tomlinson et al., 2025).

Geographic and Demographic Patterns in Adoption

While AI usage remains highest in high-income countries with advanced digital infrastructure, 2024 saw dramatic growth in low- and middle-income nations, narrowing the adoption gap (Microsoft AI Economy Institute, 2025; Chatterji et al., 2025). Attitudes vary considerably by region. Surveys show people in Asia and Latin America more likely to agree that "products and services using AI have more benefits than drawbacks"—83% in China, 70% in Mexico—while agreement is lower in Europe and Anglosphere countries like the United States (39%) and Netherlands (36%) (Maslej et al., 2025).

Language support significantly influences adoption patterns. Countries with predominant languages well-served by existing models show higher usage rates. In countries where local languages receive limited model support, users sometimes conduct conversations in English at rates disproportionate to the population's English proficiency—a pattern observed in African and Asian countries but not Europe or the Americas (Slaughter & Daepp, in prep.).

Usage purposes also vary geographically. Among early adopters, LLM use for schooling increases with GDP per capita while leisure use decreases, potentially reflecting differences in school-age populations or available leisure time (Slaughter & Daepp, in prep.).

Within the United States, workplace AI adoption varies by occupation, industry, and demographic characteristics. A 2024 survey found men slightly more likely than women to use generative AI for work (29.1% versus 23.5%) (Bick et al., 2024). Industry leaders report highest usage and confidence in IT and Procurement functions, with Technology, Telecom, Professional Services, and Finance industries leading adoption (Korst et al., 2025).

Drivers and Barriers to Organizational Adoption

Successful organizational AI adoption depends as much on employees as on leadership. The intention to use AI is influenced by social norms learned from leaders and peers across industries (Kelly et al., 2023). Workers can be reluctant to adopt top-down mandated AI products that prioritize efficiency over quality and creativity, potentially undermining traditional views of humans as core value drivers. This reluctance limits pilot program success even when technical capabilities are strong (Young et al., 2025; Sharma, 2025).

Leaders facilitate adoption through clear communication supporting AI use, demonstrating their own learning, and setting realistic expectations about AI capabilities (Carter et al., 2024; Tursunbayeva & Chalutz-Ben Gal, 2024). AI products that integrate human thinking, creativity, and expertise while amplifying their value promote adoption without raising replacement concerns (Ali et al., 2025). Organizations that create systems and incentives for employees to share how they use AI with one another—allowing insights to emerge "from the edge, not the center"—see more organic adoption (Winsor, 2024).

Employees are more likely to experiment with AI and share insights when they feel psychologically safe and trust their organizations (Tursunbayeva & Chalutz-Ben Gal, 2024). Many workers, particularly Generation X, resist tools that force conformity to a single way of working, preferring products flexible enough to accommodate personal workflows (Rozsa et al., 2023).

Top management support, customer orientation, and industry social norms drive organizational adoption intensity (Chen & Tajdini, 2024). Organizations best positioned for AI adoption are innovative, experimental, learning-oriented, supportive, and collaborative (de Bellefonds et al., 2024; Sternfels & Atsmon, 2025). However, leaders face difficulties developing top-down AI strategies due to rapid technology diffusion, constant change, the need for alignment communication, competing priorities, and the challenge of reimagining workflows (Leonardi, 2023).

A comparative case study in the Dutch public sector identified organizational inflexibility, risk intolerance, and structural separation between exploration and exploitation teams as key barriers—for example, when data science teams explore AI without operational alignment or frontline support to scale ideas (Selten & Klievink, 2024).

The Critical Role of Worker Voice in AI Design

Centering worker perspectives in AI design yields better productivity, job satisfaction, and skill development outcomes while supporting both business success and worker flourishing. This principle has deep historical roots. Research from the 1930s through the present consistently shows that when workers' expertise and perspectives inform technology design and deployment, organizations achieve more sustainable improvements in productivity and wellbeing (Trist & Bamforth, 1951; Roethlisberger & Dickson, 1939; Hackman & Oldham, 1976).

Ethnographic and human-computer interaction research demonstrates that workers adapt technology in creative ways. Participatory design approaches—where workers serve as co-designers—result in tools that better fit real workflows and foster higher adoption (Suchman, 1987; Orr, 1996; Awumey et al., 2024). Combining technical and social science research methods can create AI systems that improve worker skills and satisfaction alongside accuracy by embedding human-centric metrics, workers' values, and skill-building into design (Bucinca, 2025).

Data-driven workplace monitoring presents mixed outcomes. While telemetry and algorithmic management can boost short-term output, they often increase stress and erode trust unless workers help define what gets measured and how data gets used (Pentland, 2012; Ajunwa, 2023). Governance with worker input produces better results than monitoring imposed from above.

Building Organizational AI Maturity

AI adoption and capability-building require organizational transformation (Kemp, 2024). The Responsible AI Organizational Maturity Model (RAI-OMM) provides a roadmap for organizations advancing their responsible AI strategy and practice (Heger et al., 2025). Based on interviews and co-design sessions with 90 RAI experts and practitioners, the model identifies 24 dimensions across three categories: Organizational Foundations require leadership commitment and organization-wide infrastructure investment; Team Approach dimensions highlight cross-discipline collaboration necessity; together these enable mature RAI Practice characterized by deep integration into AI development and deployment processes.

RAI maturity requires leadership investment, aligned organizational practices, and holistic change management strategies addressing both technological and human dimensions (Duran, 2025; Wang et al., 2025). The RAI-OMM is forward-looking and best used for planning rather than evaluation, helping organizations map their current state and chart paths toward more mature practice.

Dimensions within the model are interdependent. Organizations cannot achieve mature practice in isolated areas—progress requires coordinated attention across governance, technical processes, workforce development, and cultural elements. This interdependence underscores why successful AI transformation extends far beyond technology implementation to encompass how people work together and how organizations learn.

Organizational and Individual Consequences of AI

Productivity Impacts: Time Savings and Output Quality

Surveyed ChatGPT Enterprise users attribute 40–60 minutes saved per day to AI use, though savings vary significantly by occupation and task (Chatterji et al., 2025). LLM-based estimates of time savings from Claude usage suggested 80–85% for legal and management tasks but only 20% for diagnostic image checking (Tamkin & McCrory, 2025). This heterogeneity reflects differences in how amenable various activities are to AI assistance.

In controlled evaluations, frontier LLMs approached quality parity with human experts across predominantly digital occupations in high-value sectors. OpenAI designed 1,320 tasks mimicking real work products; the top model achieved win-plus-tie shares ranging from 33–56% across industries, with low tie rates indicating clear quality differentiation (Patwardhan et al., 2025).

Analysis of Microsoft Copilot usage reveals how time allocation shifts with AI assistance. Using WorkflowView—an LLM-powered system that categorizes telemetry action sequences into high-level workflow activities—researchers found an average difference of seven minutes per accepted Copilot output among 50,000 enabled Word users over one month. Copilot use associated with a difference of 10.7 minutes in editing content and 0.6 minutes in applying themes and styles (Verma & Counts, 2025). These variations can guide more effective AI tool integration in productivity workflows.

The Workslop Problem: When AI Output Undermines Productivity

AI "workslop" refers to AI-generated work content that appears useful but lacks substance, is incomplete, or contains inaccuracies. Such content undermines productivity by forcing recipients to interpret, correct, or redo work (Niederhoffer et al., 2025; Madsen & Puyt, 2025). Workslop may explain why individual productivity gains do not consistently translate to group or organizational levels.

In a survey of 1,150 U.S. employees, 40% received workslop in the past month, estimated at 15% of content. Most slop flows between peers (40%), but also moves upward (18%) and downward (16%) in hierarchies (Niederhoffer et al., 2025). This phenomenon is part of broader "slop" dynamics reshaping markets by flooding them with low-cost, low-quality content (Tullis, 2025; Pendergrass et al., 2025).

Technical solutions remain nascent. One approach focuses on judging information utility, information quality, and style quality (Shaib et al., 2025), ideally combined with accuracy verification accessing internal data or document repositories. Employee training on AI limitations and critical evaluation skills can reduce workslop by helping people identify and correct low-value outputs before they enter workflows (Park et al., 2025; Simkute et al., 2024).

Labor Market Effects: Small Aggregate Impacts, Emerging Pressures on Early-Career Workers

Large-scale studies in Denmark and the United States find no significant AI effect on unemployment (Chen et al., 2025), working hours (Humlum & Vestergaard, 2025), or job openings (Hartley et al., 2025). Hiring of AI talent, however, has increased over 300% in the past eight years (LinkedIn, 2025).

Earnings results vary from slightly increased (Hartley et al., 2025)—especially for high AI-exposure occupations (Chen et al., 2025)—to no significant effects (Humlum & Vestergaard, 2025) to reduced salaries in high-wage occupations (Klein Teeselink, 2025). These mixed findings likely reflect heterogeneity across industries, occupations, and time periods.

Evidence of negative effects for younger workers is more consistent. Workers whose roles rely less on tacit experience are more vulnerable to automation and less shielded by firm-specific skills. Payroll data suggests employment for workers aged 22–25 in highly AI-exposed jobs fell approximately 13% compared to less-exposed roles, after testing for firm-level shocks, remote work, and sector effects (Brynjolfsson et al., 2025). Resume and job posting evidence shows hiring for junior and entry-level roles slowing in exposed occupations after firms adopt AI (Hosseini & Lichtinger, 2025; Klein Teeselink, 2025).

Declines for younger workers may be offset by growth among older workers and in less-exposed occupations, suggesting redistribution rather than net job loss. These patterns align with theory suggesting wages most likely increase when automated tasks require less expertise than other activities within the same occupation (Autor & Thompson, 2025).

Career Paths and Skill Requirements Within Occupations

AI adoption reshapes career decisions and occupational mobility. Workers using AI chatbots are more likely to switch occupations (Humlum & Vestergaard, 2025). Search intensity for apprenticeships in cognitive and language-intensive fields declined after chatbot introduction, signaling shifts in career preferences (Goller et al., 2025).

Worker-level evidence from Germany shows AI exposure changes activity mix and required skills inside occupations. Unlike robots, AI reduces non-routine abstract tasks and increases demand for high-level routine tasks like oversight and evaluation (Engberg et al., 2025; Gathmann et al., 2024). AI adoption increases complexity in augmentation-prone roles while reducing skill requirements in automation-prone roles (Chen et al., 2024).

Roles requiring AI skills are nearly twice as likely to also request analytical thinking, resilience, ethics, or digital literacy. A doubling of AI-specific job postings associates with roughly 5% higher demand for these complementary skills, while demand for easily substitutable tasks like basic data skills or translation declines slightly (Mäkelä & Stephany, 2025). Job postings requiring AI skills are growing over 70% year-over-year, extending beyond technical roles (LinkedIn, 2025a; 2025b).

Workers exposed to AI gain most from retraining focused on broad skills rather than narrow AI-specific roles. Occupations exposed to AI show strong adaptive capacity, suggesting retraining can work if job loss occurs (Hyman et al., 2025; Manning & Aguirre, 2025). However, experimental evidence suggests that while generative AI can enable non-technical workers to perform technical tasks, these gains may be temporary and dependent on continued tool use—workers lose capability to perform those tasks once access ends, indicating no lasting skill development (Wiles et al., 2024).

Online Labor Markets: Declining Demand for Writing and Design, Rising Complexity

After ChatGPT's release, automation-prone clusters on freelance platforms saw larger declines relative to manual-intensive clusters: writing jobs declined approximately 30%, software/app/web approximately 21%, and engineering approximately 10%. Image-generating AI led to approximately 17% fewer posts in graphic design and 3D modeling (Demirci et al., 2025). Some studies report rising demand for web development (Qiao et al., 2024) or complementary clusters including AI-powered chatbot development and machine learning (Teutloff et al., 2025).

The remaining automation-prone openings are more complex and slightly higher paying, but competition intensified as more applicants apply per posting (Demirci et al., 2025; Liu et al., 2025). Freelancers who adopt AI tools or shift toward complementary skills and AI-related work maintain or expand their opportunities (Qiao et al., 2024). Similar to broader labor market findings, online platform data suggests slowing demand for junior and entry-level roles in exposed occupations alongside rising value of advanced skills, human judgment, and adaptability (Teutloff et al., 2025).

Generative AI lowered the cost of producing written content, undermining the signaling value of tailored written applications. Before LLM adoption, employers paid a premium for highly customized proposals—a premium that largely disappeared afterward. Structural estimates indicate top-quintile workers are hired less often while bottom-quintile hires increase, reducing overall matching efficiency (Galdin & Silbert, 2025).

Evidence-Based Organizational Responses

Table 1: Impact and Adoption Patterns of Generative AI in the Workplace

AI Adoption Metric or Trend	Affected Group or Sector	Key Findings	Productivity or Labor Impact	Organizational Strategy Recommendations	Barriers to Success
8-fold increase in enterprise messages	Enterprise ChatGPT users	Adoption accelerated dramatically in 2024; the gender gap in usage has effectively closed.	Not in source	Prioritize employee engagement, trust, and participatory design over top-down mandates.	Workslop (AI-generated content lacking substance); cognitive deskilling risks.
40 to 60 minutes saved daily	Knowledge workers (Writing, coding, customer service)	Individual enterprise users report significant time savings and quality improvements through AI use.	Substantial individual productivity gains across digital domains.	Shift focus from individual gains to collective productivity; navigate social processes and multiple perspectives.	AI systems optimized for individuals do not automatically succeed in team settings.
40 percent of employees received workslop	U.S. employees	Workslop accounts for an estimated 15% of received content, mostly flowing between peers.	Undermines productivity by forcing recipients to interpret, correct, or redo work.	Train employees on AI limitations and critical evaluation skills.	Low-cost, low-quality content flooding markets; difficulty judging information utility.
13 percent employment decline	Early-career workers (Aged 22–25)	Workers in highly AI-exposed roles with less tacit experience face higher vulnerability to automation.	Reduction in hiring for junior and entry-level roles in exposed occupations.	Commit to reskilling and redeployment rather than layoffs; focus retraining on broad skills.	Exposure to automation due to lack of firm-specific skills and tacit experience.
30 percent decline in writing jobs	Online Freelance Platforms	Automation-prone clusters saw larger declines; graphic design posts fell 17%.	Decreased demand for easily substitutable tasks like basic data skills or translation.	Help workers shift toward complementary skills (AI-powered development) and machine learning.	Erosion of signaling value in written applications due to low-cost AI content.
Not in source	Dutch Public Sector / Organizations	Successful adoption is influenced by social norms and psychological safety.	Not in source	Use participatory design; create 'learning zones'; implement Responsible AI Maturity Models.	Organizational inflexibility; risk intolerance; structural separation between exploration and exploitation teams.

Transparent Communication and Realistic Expectation-Setting

Organizations that clearly communicate AI's role, capabilities, and limitations create foundations for effective adoption. Leaders who demonstrate their own learning journey with AI—including acknowledging uncertainties and mistakes—build psychological safety for experimentation (Carter et al., 2024). Setting realistic expectations prevents the disappointment and disengagement that follow when AI underperforms inflated promises.

Effective approaches include:

• Regular leadership communications about AI strategy, use cases, and evolving capabilities

• Public learning sessions where leaders share their AI experimentation, including what worked and what did not

• Explicit acknowledgment of AI limitations in specific contexts relevant to the organization's work

• Channels for employee questions and concerns about AI's role in their work

Organizational Example: A professional services firm instituted monthly "AI learning hours" where partners shared their AI experiments across practice areas. Early sessions featured both successes (automating routine client research) and failures (hallucinated citations in legal briefs), creating norms around transparent experimentation and collective learning. Adoption rates increased significantly after these sessions began, particularly among initially skeptical practitioners.

Participatory Design and Worker-Centered Development

Involving workers in AI design and implementation decisions produces tools that better fit real workflows while increasing adoption and satisfaction. Participatory approaches range from structured co-design sessions to grassroots bottom-up sharing of discovered use cases.

Effective approaches include:

• Worker representation on AI product selection committees to ensure frontline perspectives shape tool choices

• Structured feedback mechanisms allowing workers to report AI system failures, suggest improvements, and share successful workflows

• Internal AI use case repositories where employees document and share how they use AI, creating peer learning opportunities

• Pilot programs with iterative refinement based on worker feedback before organization-wide rollout

• Cross-functional design teams including workers from affected roles alongside technical staff

Organizational Example: A healthcare system formed interdisciplinary teams including nurses, physicians, administrative staff, and IT personnel to design AI-assisted documentation workflows. Early prototypes that looked efficient to developers proved clunky in actual clinical contexts—for example, requiring too many clicks during patient interactions. Worker feedback led to voice-activated shortcuts and ambient documentation approaches that physicians found far more natural, significantly increasing adoption.

Psychological Safety and Trust-Building

Employees experiment with AI and share insights when they feel safe making mistakes and trust organizational intentions. Organizations that punish AI-related errors or use AI monitoring in punitive ways suppress the experimentation needed to discover valuable applications.

Effective approaches include:

• Explicit "learning zones" where AI experimentation carries no performance consequences

• Transparent data governance making clear what AI systems observe, how data gets used, and who has access

• Worker input on monitoring and evaluation metrics to ensure fairness and relevance

• Commitment to reskilling and redeployment rather than layoffs when AI changes roles

• Recognition and rewards for productive AI experimentation, not just successful outcomes

Organizational Example: A financial services company established "AI sandbox" environments where analysts could experiment with AI tools on anonymized data without results affecting performance evaluations. The company committed that no one would lose their job due to AI-driven efficiency gains during a three-year transition period, with displaced workers receiving retraining for higher-value analytical work. This commitment enabled honest conversations about which tasks AI could handle and which required human judgment.

Flexible Tools Accommodating Personal Workflows

Many workers, particularly experienced professionals, have developed personal workflows optimized over years. Rigid AI tools requiring conformity to a single process face resistance. Flexible systems that adapt to varied working styles see higher adoption.

Effective approaches include:

• Customizable AI interfaces allowing users to adjust interaction patterns, output formats, and level of assistance

• Multiple interaction modalities (chat, command line, API, GUI) serving different user preferences and contexts

• Configurable autonomy levels enabling users to adjust how much initiative AI takes versus waiting for explicit direction

• Integration with existing tools rather than requiring workflow disruption to use AI

Organizational Example: A software development company deployed coding assistants with configurable settings allowing developers to choose between "suggestion mode" (AI offers completion options the developer accepts or rejects), "collaborative mode" (iterative back-and-forth refinement), and "autonomous mode" (AI completes specified tasks end-to-end with human review). Different developers gravitated toward different modes based on task complexity, domain familiarity, and personal preference, with overall adoption higher than earlier one-size-fits-all pilot programs.

Responsible AI Governance and Ethical Frameworks

Building organizational capacity for responsible AI development and deployment prevents harms while building stakeholder trust. Mature responsible AI practice requires more than policies—it demands leadership investment, cross-functional collaboration, and integration into product development processes.

Effective approaches include:

• Dedicated responsible AI teams with authority to require changes before deployment

• Impact assessments examining potential harms across stakeholder groups before deploying AI systems

• Regular audits of deployed AI systems checking for drift, bias, and unintended consequences

• Transparent documentation of AI system capabilities, limitations, and appropriate use cases

• Mechanisms for recourse when AI systems produce harmful outputs

• Investment in organizational learning about responsible AI principles and practices

Organizational Example: A technology company established a Responsible AI Office reporting directly to the CEO, with authority to delay product launches pending resolution of identified risks. The office developed streamlined impact assessment templates tailored to different AI application categories, making responsible AI review integral to the product development lifecycle rather than a bureaucratic obstacle. Teams received training in fairness, transparency, privacy, and reliability principles, with responsible AI metrics incorporated into performance evaluation for product managers and engineers.

Building Long-Term AI Capability and Governance

Continuous Learning and Skill Development

Organizations that treat AI capability-building as ongoing learning rather than one-time training create adaptive workforces prepared for rapid technological change. This requires investment in formal training, peer learning opportunities, and time for experimentation.

Effective approaches include:

• Regular skill assessments identifying AI-related capability gaps across roles and levels

• Diverse learning modalities including formal courses, peer learning communities, hands-on experimentation, and expert-led workshops

• Career pathways incorporating AI skills so workers see capability development as advancement rather than just avoiding displacement

• Dedicated time for learning rather than expecting skill development on top of full workloads

• Internal expertise sharing platforms where workers document successful AI applications and lessons learned

Organizations must balance building general AI literacy (understanding capabilities, limitations, appropriate use) with role-specific technical skills. A customer service representative needs different AI competencies than a data scientist, but both benefit from understanding how AI systems work and where they struggle.

Distributed Leadership and Decision-Making Authority

Centralized AI strategies imposed from above often miss context-specific opportunities and constraints visible only at operational levels. Distributing decision-making authority while maintaining appropriate guardrails enables innovation "from the edge."

Effective approaches include:

• Empowering teams to choose AI tools within defined parameters rather than mandating single solutions

• Budget allocation for team-level AI experimentation without requiring extensive justification

• Lightweight approval processes balancing autonomy with responsible oversight

• Cross-team sharing mechanisms propagating successful innovations discovered at operational levels

• Clear escalation paths for questions exceeding local authority

This approach does not mean abandoning governance. Organizations still need enterprise-wide standards for security, privacy, ethics, and interoperability. The balance lies in defining what requires central control versus what benefits from local discretion.

Purpose-Driven AI Integration Supporting Meaningful Work

Workers engage more deeply with AI when they understand how it serves broader organizational purposes and preserves aspects of work they find meaningful. AI implemented purely for efficiency often meets resistance; AI positioned as enabling workers to focus on higher-value contributions tends to see stronger adoption.

Effective approaches include:

• Explicit connection between AI initiatives and organizational mission, showing how AI advances core purpose

• Worker input on which tasks to automate or augment, respecting preferences about maintaining connection to certain activities

• Recognition systems valuing judgment, creativity, and relational work that AI cannot easily replicate

• Career development emphasizing uniquely human capabilities like empathy, ethical reasoning, and complex problem-solving

• Regular dialogue about work's meaning and purpose as AI changes daily activities

Organizations risk eroding motivation and psychological wellbeing when AI removes aspects of work that provided autonomy, recognition, or connection—key elements of meaningful work (Bailey et al., 2019). Thoughtful AI integration preserves or enhances these dimensions rather than optimizing solely for measurable productivity.

Data and Model Stewardship

Organizations deploying AI systems have ongoing responsibilities for data governance, model maintenance, and continuous improvement. These stewardship functions require dedicated resources and clear accountability.

Effective approaches include:

• Data quality monitoring ensuring training and operational data meet standards for accuracy, representativeness, and timeliness

• Model performance tracking detecting drift, degradation, or emerging failure modes

• Version control and change management maintaining clear records of model updates and their rationales

• Stakeholder feedback loops incorporating user experiences and concerns into model refinement

• Decommissioning processes for retiring AI systems that no longer serve intended purposes or create unacceptable risks

Organizations must resist treating AI systems as "set and forget" deployments. Continuous stewardship prevents the gradual emergence of problems that undermine trust and effectiveness.

Adaptive Governance Structures

The rapid pace of AI capability advancement demands governance frameworks that can evolve alongside technology. Static policies written for today's capabilities may prove inadequate or overly restrictive as systems improve.

Effective approaches include:

• Regular policy review cycles triggered by capability milestones or emerging use cases

• Scenario planning exploring potential implications of advancing AI capabilities

• External advisory input bringing diverse perspectives on emerging AI governance challenges

• Flexible frameworks stating principles and goals rather than rigid procedural requirements

• Experimentation mechanisms allowing controlled trials of new AI applications with defined learning objectives

Organizations should view governance as enabling rather than constraining AI value creation. Well-designed governance clarifies boundaries within which innovation can proceed confidently while preventing harmful applications.

Conclusion

The evidence synthesized in this report reveals AI's transformative potential alongside significant implementation challenges. Individual productivity gains are real and substantial—workers save time, produce higher-quality outputs, and access capabilities previously requiring extensive expertise. These individual benefits do not yet consistently translate to team and organizational levels, where social dynamics, coordination requirements, and collective intelligence present distinct challenges.

Successful organizational AI adoption depends critically on centering worker voice, building trust, enabling flexible workflows, and viewing technology implementation as sociotechnical transformation rather than pure technical deployment. Organizations that mandate AI use from above, ignore worker concerns, or optimize narrowly for efficiency metrics struggle with resistance and limited adoption. Those that invest in participatory design, transparent communication, psychological safety, and continuous learning create conditions for sustainable productivity gains.

Labor market impacts remain modest in aggregate but show concerning patterns of pressure on early-career workers in exposed occupations. Whether these early signals foreshadow broader disruption or represent transitional friction depends substantially on organizational and policy choices not yet made. The path forward is not technologically determined—human decisions about whether to prioritize automation versus augmentation, how to structure work, and how to distribute AI-enabled gains will shape outcomes.

Design challenges around human-AI collaboration demand sustained attention. Establishing common ground, managing proactivity, supporting cognitive engagement rather than passive reliance, and building AI systems that understand team dynamics all require advances beyond current capabilities. The transition from reactive AI assistants to proactive team members raises fundamental questions about turn-taking, shared understanding, long-horizon alignment, and appropriate evaluation metrics.

Critical risks warrant attention: cognitive deskilling when AI handles tasks that previously built expertise; psychological dependence on AI companions; erosion of social connection when AI mediates too much human interaction; concentration of economic gains absent deliberate policy intervention; and the potential for AI-accelerated research to create substitution dynamics that diminish human contributions. These risks are not inevitable—thoughtful design and governance can mitigate them—but require explicit prioritization alongside productivity metrics.

The future of work with AI will be shaped not by technology alone but by conscious choices about how we design systems, structure organizations, develop policy frameworks, and define success. The transition from individual to collective productivity represents not just a technical scaling challenge but an opportunity to reimagine work in ways that expand human capability, preserve meaningful engagement, and distribute benefits broadly. Achieving that future requires moving beyond viewing AI as a tool that replaces human labor toward understanding it as a medium through which human intelligence can be amplified, connected, and applied to challenges previously beyond reach.

Research Infographic

From Individual Gains to Collective Intelligence Slide Deck

References

1. Ajunwa, I. (2023). The quantified worker. Cambridge University Press.

2. Ali, D., Dennison, M., & Rajpurkar, P. (2025). AI adoption across mission-driven organizations. arXiv preprint arXiv:2501.00000.

3. Autor, D., & Thompson, N. (2025). Expertise. National Bureau of Economic Research Working Paper.

4. Awumey, E., Birchfield, S., & Kamar, E. (2024). A systematic review of biometric monitoring in the workplace: Analyzing sociotechnical harms in development, deployment and use. ACM Conference on Fairness, Accountability, and Transparency.

5. Bailey, C., Madden, A., Alfes, K., & Fletcher, L. (2019). A review of the empirical literature on meaningful work: Progress and research agenda. Human Resource Development Review, 18(1), 83–113.

6. Bick, A., Blandin, A., & Mertens, K. (2024). The rapid adoption of generative AI. National Bureau of Economic Research Working Paper.

7. Brynjolfsson, E. (2022). The Turing trap: The promise and peril of human-like artificial intelligence. Daedalus, 151(2), 272–287.

8. Brynjolfsson, E., Horton, J., Ozimek, A., Rock, D., Sharma, G., & TuYe, H. (2025). Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence. Working Paper.

9. Bucinca, Z. (2025). Worker-centric AI for decision-support. Harvard University Dissertation.

10. Carter, J., Edmondson, A., & Pisano, G. (2024). To succeed with AI, adopt a beginner's mindset. Harvard Business Review.

11. Chatterji, A., Li, D., & Raymond, L. (2025). How people use ChatGPT. National Bureau of Economic Research Working Paper.

12. Chen, D., Saridakis, G., Goel, S., & Kontonikas, A. (2025). The (short-term) effects of large language models on unemployment and earnings. arXiv preprint arXiv:2501.00000.

13. Chen, J., & Tajdini, S. (2024). A moderated model of artificial intelligence adoption in firms and its effects on their performance. Information Technology & Management, 25, 291–310.

14. Chen, W. X., Huang, Z., & Kumar, A. (2024). Displacement or complementarity? The labor market impact of generative AI. Harvard Business School Working Paper.

15. de Bellefonds, N., Chin, A., Chinn, D., Hill, A., & Singla, A. (2024). Where's the value in AI? Boston Consulting Group Report.

16. Dell'Acqua, F., McFowland, E., Mollick, E., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., ... & Lakhani, K. (2025). The cybernetic teammate: A field experiment on generative AI reshaping teamwork and expertise. Harvard Business School Working Paper.

17. Demirci, O., Kolaghassi, R., & Demirer, M. (2025). Who is AI replacing? The impact of generative AI on online freelancing platforms. Management Science, Forthcoming.

18. Duran, J., Wang, A., Kocielnik, R., Nushi, B., Kamar, E., & Andersen, E. (2025). RAI advocacy: Communicative strategies for advancing responsible AI in large technology companies. AAAI/ACM Conference on AI, Ethics, and Society.

19. Engberg, E., Dengler, K., & Matthes, B. (2025). Artificial intelligence, tasks, skills and wages: Worker-level evidence from Germany. Research Policy, 54(1), 104898.

20. Galdin, J., & Silbert, S. (2025). Making talk cheap: Generative AI and labor market signaling. Working Paper.

21. Gathmann, C., Helm, I., & Schönberg, U. (2024). AI, task changes in jobs, and worker reallocation. CESifo Working Paper.

22. Goller, D., Heim, S., & Kuom, D. (2025). This time it's different – Generative artificial intelligence and occupational choice. Labour Economics, 85, 102574.

23. Hackman, J. R., & Oldham, G. R. (1976). Motivation through the design of work: Test of a theory. Organizational Behavior and Human Performance, 16(2), 250–279.

24. Handa, K., Kreps, S., McCain, R., Schaekermann, M., Veselovsky, V., Schoelkopf, H., ... & Tetlock, P. (2025). Which economic tasks are performed with AI? Evidence from millions of Claude conversations. arXiv preprint arXiv:2501.00000.

25. Hartley, J., Huang, K., Kaplan, G., & Pivonka, M. (2025). The labor market effects of generative artificial intelligence. SSRN Working Paper.

26. Heger, A., Wang, A., Duran, J., Andersen, E., Kocielnik, R., Kamar, E., & Nushi, B. (2025). Towards a responsible AI organizational maturity model. ACM Conference on Computer-Supported Cooperative Work and Social Computing.

27. Hosseini, S. M., & Lichtinger, G. (2025). Generative AI as seniority-biased technological change: Evidence from U.S. résumé and job posting data. SSRN Working Paper.

28. Humlum, A., & Vestergaard, E. (2025). Large language models, small labor market effects. National Bureau of Economic Research Working Paper.

29. Hyman, B., Martinez, J., & Hadavand, A. (2025). How retrainable are AI-exposed workers? Working Paper.

30. Kelly, S., Kaye, S. A., & Oviedo-Trespalacios, O. (2023). What factors contribute to the acceptance of artificial intelligence? A systematic review. Telematics and Informatics, 77, 101925.

31. Kemp, A. (2024). Competitive advantage through artificial intelligence: Toward a theory of situated AI. Academy of Management Review, 49(3), 441–464.

32. Klein Teeselink, B. (2025). Generative AI and labor market outcomes: Evidence from the United Kingdom. SSRN Working Paper.

33. Korst, J., Ladd, T., & Loftus, C. (2025). Accountable acceleration: Gen AI fast-tracks into the enterprise. Wharton Human-AI Research and GBK Collective Report.

34. Leonardi, P. (2023). Helping employees succeed with generative AI. Harvard Business Review.

35. LinkedIn. (2025). AI labor market update: Tracking AI adoption and skills in the U.S. economy. LinkedIn Economic Graph Report.

36. LinkedIn. (2025a). Work change report: AI is coming to work. LinkedIn Research.

37. LinkedIn. (2025b). LinkedIn skills on the rise 2025: The 15 fastest-growing skills in the U.S. LinkedIn Economic Graph.

38. Liu, J., Yuan, H., Cosley, D., & Guo, Y. J. (2025). "Generate" the future of work through AI: Empirical evidence from online labor markets. Under Review.

39. Madsen, D. Ø., & Puyt, R. W. (2025). The 7Vs of AI slop: A typology of generative waste. SSRN Working Paper.

40. Mäkelä, E., & Stephany, F. (2025). Complement or substitute? How AI increases the demand for human skills. Working Paper.

41. Manning, S., & Aguirre, T. (2025). How adaptable are American workers to AI-induced job displacement? National Bureau of Economic Research Working Paper.

42. Maslej, N., Fattorini, L., Brynjolfsson, E., Etchemendy, J., Ligett, K., Lyons, T., ... & Perrault, R. (2025). The AI Index 2025 annual report. AI Index Steering Committee, Institute for Human-Centered AI, Stanford University.

43. Microsoft AI Economy Institute. (2025). AI diffusion report: Where AI is most used, developed, and built. Microsoft Corporation.

44. Niederhoffer, K., Luca, M., & Schwartz, J. (2025). AI-generated "workslop" is destroying productivity. Harvard Business Review.

45. Orr, J. (1996). Talking about machines: An ethnography of a modern job. Cornell University Press.

46. Park, J., Barber, L., Cho, E., & Wang, Y. (2025). Attitudes towards artificial intelligence at work: Scale development and validation. Journal of Occupational and Organizational Psychology, 98(1), 195–218.

47. Patwardhan, T., Li, Y., Li, Z., Tamkin, A., & Burns, K. (2025). GDPval: Evaluating AI model performance on real world economically valuable tasks. arXiv preprint arXiv:2501.00000.

48. Pendergrass, W., Adams, T., & Sample, C. (2025). A strategic cycle of slop: Understanding the commodification of AI feculence and its place in the attention economy. Issues in Information Systems, 26(1).

49. Pentland, A. (2012). The new science of building great teams. Harvard Business Review, 90(4), 60–70.

50. Qiao, D., Luo, X., & Nan, G. (2024). AI and freelancers: Has the inflection point arrived? International Conference on Information Systems.

51. Roethlisberger, F., & Dickson, W. (1939). Management and the worker. Harvard University Press.

52. Rozsa, Z., Formánková, S., & Majerová, J. (2023). Job crafting and sustainable work performance: A systematic literature review. Equilibrium. Quarterly Journal of Economics and Economic Policy, 18(3), 717–750.

53. Schmutz, J. B., Knobel, S. E. J., Zahner, N., & Manser, T. (2024). AI-teaming: Redefining collaboration in the digital era. Current Opinion in Psychology, 59, 101878.

54. Selten, F., & Klievink, B. (2024). Organizing public sector AI adoption: Navigating between separation and integration. Government Information Quarterly, 41(2), 101909.

55. Shaib, C., de Wynter, A., & Jiang, Z. (2025). Measuring AI "slop" in text. arXiv preprint arXiv:2501.00000.

56. Sharma, R. (2025). The impact of AI-generated content on human creativity and original thought: A psychological study. American Psychological Association Presentation.

57. Simkute, A., Daepp, M., Kuehl, K., & Teevan, J. (2024). Ironies of generative AI: Understanding and mitigating productivity loss in human-AI interaction. International Journal of Human-Computer Interaction, 40(19), 4927–4945.

58. Slaughter, I., & Daepp, M. (in prep.). Second-level digital divides in usage of a generative AI chatbot.

59. Sternfels, B., & Atsmon, Y. (2025). The learning organization: How to accelerate AI adoption. McKinsey & Company.

60. Suchman, L. (1987). Plans and situated actions: The problem of human-machine communication. Cambridge University Press.

61. Tamkin, A., & McCrory, P. (2025). Estimating productivity gains from Claude conversations. Anthropic Research.

62. Teutloff, O., Kampkötter, P., & Wilmers, N. (2025). Winners and losers of generative AI: Early evidence of shifts in freelancer demand. Journal of Economic Behavior & Organization, 219, 1–14.

63. Tomlinson, K., Daepp, M., Linden, S., Cheng, S., & Hecht, B. (2025). Working with AI: Measuring the applicability of generative AI to occupations. arXiv preprint arXiv:2501.00000.

64. Trist, E., & Bamforth, K. (1951). Some social and psychological consequences of the longwall method of coal-getting. Human Relations, 4(1), 3–38.

65. Tullis, J. (2025). Sifting through the slop: How generative AI created a market for lemons for text-based works. SSRN Working Paper.

66. Tursunbayeva, A., & Chalutz-Ben Gal, H. (2024). Adoption of artificial intelligence: A TOP framework-based checklist for digital leaders. Business Horizons, 67(6), 687–701.

67. Verma, G., & Counts, S. (2025). WorkflowView: Abstracting activity logs with LLMs for interpretable and actionable insights. Under Review.

68. Wang, A., Kocielnik, R., Duran, J., Andersen, E., Kamar, E., & Nushi, B. (2024). Strategies for increasing corporate responsible AI prioritization. AAAI/ACM Conference on AI, Ethics, and Society.

69. Wiles, E., Zhang, M., & Autor, D. (2024). GenAI as an exoskeleton: Experimental evidence on knowledge workers using GenAI on new skills. SSRN Working Paper.

70. Winsor, J. (2024). How to be systematic about adopting AI at your company. Harvard Business Review.

71. Young, J., Rozsa, A., & Beers, S. (2025). The future of enterprise. Internal Microsoft Report.

---

Jonathan H. Westover, PhD is Chief Research Officer (Nexus Institute for Work and AI); Associate Dean and Director of HR Academic Programs (WGU); Professor, Organizational Leadership (UVU); OD/HR/Leadership Consultant (Human Capital Innovations). Read Jonathan Westover's executive profile here.

---

Suggested Citation: Westover, J. H. (2026). The Future of Work with AI: Moving from Individual Gains to Collective Intelligence. Human Capital Leadership Review, 32(1). doi.org/10.70175/hclreview.2020.32.1.3