This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to papers

Large Language Models, Small Labor Market Effects

Humlum, Vestergaard

2025NBER Working Paper Series25 citations
Observational labor marketCausalTheoretical model
LLM / Generative AIJunior / entry-levelGenderTraining / upskillingAugmentation vs. substitution
DOI: 10.3386/w33777
Abstract

We examine the labor market effects of AI chatbots using two large-scale adoption surveys (late 2023 and 2024) covering 11 exposed occupations (25,000 workers, 7,000 workplaces), linked to matched employer-employee data in Denmark.AI chatbots are now widespread-most employers encourage their use, many deploy in-house models, and training initiatives are common.These firm-led investments boost adoption, narrow demographic gaps in take-up, enhance workplace utility, and create new job tasks.Yet, despite substantial investments, economic impacts remain minimal.Using difference-in-differences and employer policies as quasi-experimental variation, we estimate precise zeros: AI chatbots have had no significant impact on earnings or recorded hours in any occupation, with confidence intervals ruling out effects larger than 1%.Modest productivity gains (average time savings of 3%), combined with weak wage pass-through, help explain these limited labor market effects.Our findings challenge narratives of imminent labor market transformation due to Generative AI.

Summary

Humlum and Vestergaard link large-scale representative surveys on AI chatbot adoption (25,000 workers, 7,000 workplaces) to Danish administrative labor records to study early labor market impacts of generative AI using difference-in-differences.

Main Finding

AI chatbots show precise null effects on earnings and hours at both worker and workplace levels two years after ChatGPT launch, with confidence intervals ruling out effects larger than 2%, despite widespread adoption and employer investment.

Primary Datasets

Two surveys (25K workers, 7K workplaces); Danish admin data

Secondary Datasets

Danish Population Register (BEF); Personal Wealth Register (FORMPERS); Firm Statistics Register (FIRM)

Key Methods
Difference-in-differences using ChatGPT launch (Nov 2022) as event, comparing adopters vs non-adopters and encouraged vs non-encouraged workplaces, linking survey responses to administrative matched employer-employee data.
Sample Period
2023-2024
Geographic Coverage
Denmark
Sample Size
25,000 survey respondents linked to administrative data; 11 AI-exposed occupations covering ~20% of Danish employment
Level of Analysis
Individual, Firm, Occupation
Occupation Classification
ISCO-based occupation codes covering 11 exposed occupations
Industry Classification
None
Notes
NBER WP 33777; precise null effects on earnings and hours; confidence intervals rule out effects >1%; Non-US paper; included for cross-national comparison [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025. [Claude classification]: Paper uses ChatGPT launch (Nov 2022) as natural experiment timing. Finds null effects despite: (1) 43% of workers in encouraged workplaces, (2) 69% ever-use rate among adopters, (3) employers investing in enterprise chatbots (38%) and training (30%). Survey includes detailed measures of employer initiatives (encouragement, enterprise chatbots, training). Adopters show 4% increase in occupational mobility (FTE in latest occupation) but no net earnings/hours gains. Creates novel link between self-reported adoption/benefits and third-party administrative outcomes. Replicates Brynjolfsson et al. (2025) finding of early-career job declines in exposed occupations but shows decline NOT driven by adopting firms. Gender gap in adoption (13pp) reduced by employer training. Uses Empirical Bayes shrinkage for workplace-level measures. Includes free-text classification of new AI-related tasks (50-95% of new tasks AI-related). Theoretical framework models chatbot adoption as Roy-style selection with employer initiatives shifting costs/benefits. Working paper revised October 2025.