AI Workforce Transformation Wiki

AI-mediated work experience

AI-mediated work experience is the integrative domain covering how AI changes the lived and motivational quality of work. It is not a unitary construct or another broad engagement label. It connects distinct resources, demands, appraisals, behaviors, and outcomes—including autonomy, competence, identity, workload, learning, meaning, control, affect, and social texture—so the wiki can examine how they combine inside the same workflow.

The focal question is configurational: when AI enters a workflow, which parts of work become more supportive, demanding, agentic, developmental, meaningful, or socially sustaining, and which become less so? A system can be a cognitive resource while increasing verification load, improve short-run affect while weakening human sociality, or raise performance while eroding competence and professional identity. Those effects should be separated analytically before they are synthesized as an overall change in work experience.

Analytic dimensions

Resources, demands, and workload

Work characteristics shape whether AI can become useful support. 2026-reich-work-design-multidimensional-ai-threat found that skill variety and autonomy were the most consistent positive predictors of workplace AI adoption, while different threat appraisals related differently to depth of use. 2026-lee-making-invisible-visible gives the corresponding workflow account: workers valued assistance that preserved meaningful control, learning, collaboration, and accountability, while AI that took the wrong work could leave them with tedious backstage tasks even when it technically improved efficiency.

Resource value cannot be inferred from fluent output or faster task completion. cognitive-load-in-ai-assisted-work shows how useful AI-generated content can coexist with extraneous interaction burden, while digital-work-strain distinguishes load, invasion, complexity, insecurity, uncertainty, support, autonomy, and recovery conditions. 2025-klonek-ai-work-stress further associates greater AI control with higher stress while finding lower stress around human–AI transition and action processes. The child topics carry the construct-specific evidence; the parent-level claim is that net work experience depends on the full resource–demand configuration.

Agency, competence, and identity

AI can improve autonomy while simultaneously weakening competence or authorship. 2026-ehsan-future-workers found that efficiency gains in radiation oncology coexisted with intuition rust, reduced manual practice, and anxiety about becoming a bystander in professional work. 2026-edwards-developer-productivity reports the same tension among senior developers: AI could operate as a nonjudgmental mentor and strategic collaborator while raising productivity baselines and creating uncertainty about foundational skill, authorship, and professional value.

Desired agency also depends on which work remains human. 2025-shao-future-work-ai-agents found broad interest in automating low-value repetitive tasks but substantial occupational preference for equal partnership or greater human involvement. ai-mediated-choice-and-identity carries the adjacent transfer evidence that delegated AI choices can preserve apparent preference compatibility while narrowing distinctiveness and exploration. The work-design implication is that agency cannot be assessed only through whether workers accept an AI tool; it also depends on whether the remaining role preserves meaningful choice, practice, identity, and control.

2026-ghosh-new-social-image makes the configuration more specific: in a small 2 × 2 × 2 video-vignette experiment, lower AI competence or proactivity generally produced more favorable perceived ownership, meaningfulness, satisfaction, role dynamics, and affect. This is early scenario-based evidence, not a verdict that better AI makes work worse. It does show why system performance and initiative should be treated as work-experience inputs: they can alter the perceived social value of the human contribution before any formal role redesign occurs.

Social and relational experience

AI changes the social experience of work as well as the human–machine interface. 2025-brynjolfsson-generative-ai-at-work observed a more positive customer interaction climate alongside productivity gains in customer support. 2025-dellacqua-super-mario-ai found that direct pairing with an automated agent reduced trust and effort even when the system performed well, while 2026-dellacqua-cybernetic-teammate found higher positive and lower negative emotion after AI-supported knowledge work. These results are not contradictory by default: short-run affect, trust in the arrangement, effort, and durable relatedness are different outcomes.

AI can also reorganize human relationships around the work. 2025-yun-generative-ai-knowledge-work shows that product managers wanted relief from fragmented information work but worried about losing source contact and collaboration to an agreeable tool. 2026-buechsenschuss-genai-organizational-networks found higher satisfaction with knowledge access and greater collaboration and knowledge-sharing centrality after access to a grounded assistant. Greater centrality may be empowering or burdensome depending on whether easier knowledge flow also produces more requests, coordination fatigue, or availability pressure.

Motivation and proactive adaptation

2026-vuong-genai-job-performance places job crafting and work engagement in a proposed JD-R pathway between GenAI usage and performance. The cross-sectional results cannot establish that GenAI caused better experience, but they clarify the nomological network: AI use, enacted resource functions, proactive work reshaping, motivational state, and performance are related but noninterchangeable. The parent page therefore integrates these channels without turning them into a single omnibus outcome.

2026-bakker-artificial-intelligence-job-design supplies an explicit dynamic JD-R account for this parent domain. The review argues that AI adoption can increase or decrease both demands and resources, while AI-readiness resources determine whether a motivational or health-impairment pathway is more likely. Its more consequential claim is recursive: resource-rich jobs may let workers build AI self-efficacy, learning, and engagement through use, while constrained, monitored work may turn AI into a demand that lowers later readiness. This is a useful inequality hypothesis for AI-mediated work experience, not evidence that gain and loss spirals have already been established in workplace AI settings.

Integrative design claim

The appropriate unit of intervention is the complete sociotechnical workflow. Organizations should assess not only whether AI improves output, but how it reallocates control, cognitive effort, verification, learning opportunities, authorship, social contact, recovery burden, and accountability. An apparently positive result in one dimension does not settle the overall experience claim. The design task is to make the trade-offs visible and decide which configuration of resources, demands, agency, capability, and sociality the work should sustain.

Connections

  • Relates to work-redesign because task allocation, control rights, verification, performance criteria, and learning opportunities determine the resulting experience.
  • Relates to employee-engagement because engagement is one possible motivational outcome, not a substitute for measuring distinct experience dimensions.
  • Relates to ai-adoption and ai-readiness because sustained use depends partly on whether workers experience AI-enabled work as supportive, threatening, developmental, governable, and worth doing.
  • Relates to responsible-ai-deployment because autonomy, competence, privacy, psychological safety, identity, workload, and felt control are deployment outcomes rather than implementation side notes.
  • Relates to ai-supported-knowledge-synthesis because source contact, context, collaboration, and accountable judgment shape the experience of AI-assisted knowledge work.

Contradictions & open questions

  • Positive productivity, quality, or adoption evidence does not establish improved work experience. The wiki needs studies that measure output and multiple experience dimensions in the same design.
  • Short-run positive emotion can coexist with lower trust, effort, relatedness, or long-run learning. Evidence should preserve those distinctions rather than aggregate them into satisfaction.
  • Greater autonomy over task execution can coexist with raised performance expectations, verification burden, or reduced control over the pace and evaluation of work.
  • AI may operate as a resource at the point of generation while shifting demands into correction, traceability, documentation, and accountability downstream. Field studies need to estimate that net balance.
  • The strongest causal evidence remains short-horizon, while much of the workplace evidence is cross-sectional, qualitative, or indirect. Longitudinal and intervention research is still needed on autonomy, competence, relatedness, identity, workload, learning, and recovery.
  • Experience is unlikely to move uniformly across workers. Occupation, expertise, task risk, desired agency, accessibility needs, managerial expectations, and governance design are plausible boundary conditions that remain unevenly tested.

Graph View

Backlinks