2026年未来工作报告：生成式AI——工具、伙伴还是责任风险？

Generative AI:Tool, Colleague, or Liability? EXECUTIVE SUMMARY Artificial intelligence (AI) isembedded in daily work acrossmost organizations. As adoptionbecomes the norm, the challenge The LexisNexis 2026 Future ofWork report asks a critical question: Is the development of AIgovernance, training, andeducation keeping pace withhow quickly genAI is being Recent studies show leadersbelieve they are ahead of thecurve in their AI journey. Yet,our research among 1,400professionals across over 20 Increasing Governance Risk Whether organizational leaders like it or not,generative AI (genAI) adoption is growing. Whilehalf of professionals surveyed report at leastfrequent genAI use, the infrastructure to supportthis hasn’t kept pace. Specifically, 53% have usedgenAI without formal approval, 28% report their The Governance Crisis 55%pay forown tools Training Itself Isn’t Enough to Overconfidence as a Liability Most (64%) professionals claim to be very orextremely confident using genAI responsibly.These high confidence levels, combined with lowunderstanding of risks, can create significantliability risk. Employees trust genAI outcomes but While 82% of professionals receive some form oftraining (up from 72% in 2025), training withoutadequate tools may be accelerating shadow AI*adoption. Those with mandatory training showhigher unauthorized usage rates than those with Professionals are using genAI more than ever, butmany don’t fully comprehend the data sources,logic, or ethical safeguards behind the tools on 74% of professionals withmandatory training usedgenAI at work without The professional workforce as a whole appearsready and eager to adopt genAI, but leaderswithin these organizations simply haven’t kept The defining shift of 2026 isthe evolution from asking“Can genAI do this?”to“Should we trust how it does TABLE OF CONTENTS Executive Summary............................................................................................22025 Versus 2026:Capability to Credibility.................................................5Key Finding 1:The Governance Crisis:Leadership’s Biggest Blind Spot....................................................................7Key Finding 2:The Overconfidence Risk: Are Employeesas Ready as They Think Are? Probably Not.............................................11Key Finding 3:Training Builds Confidence,But Access Remains Uneven.......................................................................14Appendix.............................................................................................................20Explore Industry-Specific Insights and Next Steps...............................20Methodology..................................................................................................20Survey Design and Measures......................................................................21 2025 VERSUS 2026:CAPABILITY TO Workplace Adoption Enthusiasm for genAI has softened year over year, with an 18-point decline from 2025 to 2026in those open to its use. While a majority remain receptive to its benefits, increased awarenessof how models are trained and how data is stored or reused has led to greater scrutiny of genAI Confidence Levels The 2025 report examined respondents’ confidence levels in genAI’s capabilities. As usagehas increased over time, safety and accuracy concerns rise with it. This year, the survey’s focusreflected increasing emphasis on governance and ethical application. Respondents were asked to not confident in usinggenAI responsibly rely on genAI to takeover entire tasks Technical Understanding Technical understanding of genAI fundamentals has grown significantly year over year, withknowledge of how large language models (LLMs) work jumping from 64% in 2025 to 77% in 2026, Familiarity with genAI writing tools Training While the 2025 survey measured how often employees receive genAI training and the 2026 surveyexamined the formats used to train employees on genAI tools and their use, both sets of resultspoint to continued variation in how organizations approach genAI workforce enablement. Acrossboth years, a meaningful proportion of organizations offered structured or ongoing employee KEY FINDING 1 The Governance Crisis:Leadership’s Biggest Blind Spot The Shadow AI Problem Many organizational leaders underestimatethe use of shadow AI* among their employees.Employees are not waiting for their company toprovide them with the formal genAI tools theyneed for productivity. As a result, a majority We need to limit its use andavailability until it has beenproperly vetted and deemedtrustworthy… and does not Response fromLexisNexis Future of Work T he industries with professionals stating the lowest confidence intheir abilities to safely use genAI are PR, media, and communications(only32%reporting confidence). Adopting its (genAI) usehas been a huge time suck,particularly since it really can’tdo what we expect, need, or The R