GenAI鸿沟：2025年AI商业现状

The GenAI DivideSTATE OF AI INBUSINESS 2025 MIT NANDA Aditya ChallapallyChris PeaseRamesh RaskarPradyumnaChariJuly 2025 NOTES Preliminary Findings from AI Implementation Research from Project NANDAReviewers:Pradyumna Chari, Project NANDAResearch Period:January–June 2025Methodology:This report is based on a multi-method research design that includesa systematic review of over 300 publicly disclosed AI initiatives, structuredinterviews with representatives from 52 organizations, and survey responses from153 senior leaders collected across four major industry conferences.Disclaimer:The views expressed in this report are solely those of the authors andreviewers and do not reflect the positions of any affiliated employers.Confidentiality Note:All company-specific data and quotes have beenanonymized to maintain compliance with corporate disclosure policies andconfidentiality agreements, ensure neutrality, and prevent any perception ofcommercial advancement or opinion. 1CONTENTS 1.Executive Summary2.The Wrong Side of the GenAI Divide: High Adoption, Low Transformation3.Why Pilots Stall: The Learning Gap Behind the Divide4.Crossing the GenAI Divide: How the Best Builders Succeed5.Crossing the GenAI Divide: How the Best Buyers Succeed6.Conclusion: Bridging the GenAI Divide 2EXECUTIVESUMMARY Despite $30–40 billion in enterprise investmentinto GenAI,this report uncovers a surprisingresult in that95% of organizations are getting zero return.The outcomes are so starklydivided across both buyers (enterprises, mid-market, SMBs) and builders (startups, vendors,consultancies) that we call it the GenAI Divide.Just 5% of integrated AI pilots are extractingmillions in value, while the vast majority remain stuck with no measurable P&L impact.Thisdividedoes not seem to bedriven by model quality or regulation, but seems to bedetermined by approach. Tools like ChatGPT and Copilot are widely adopted. Over 80 percent of organizations haveexplored or piloted them, and nearly 40 percent report deployment. But these toolsprimarily enhance individual productivity, not P&L performance. Meanwhile, enterprise-grade systems,custom or vendor-sold,are being quietly rejected. Sixty percent oforganizations evaluated such tools, but only 20 percent reached pilot stage and just 5percent reached production. Most fail due to brittle workflows, lack of contextual learning,and misalignment with day-to-day operations. From our interviews, surveys, and analysis of 300 public implementations, four patternsemerged that define the GenAI Divide: •Limited disruption: Only 2 of 8 major sectors show meaningful structural change•Enterprise paradox: Big firms lead in pilot volume but lag in scale-up•Investment bias: Budgets favor visible, top-line functions over high-ROI back office•Implementation advantage: External partnerships see twice the success rate ofinternal builds The core barrier to scaling is not infrastructure, regulation, or talent. It is learning. MostGenAI systems do not retain feedback, adapt to context, or improve over time. A small group of vendors and buyers are achieving faster progress by addressing theselimitations directly. Buyers who succeed demand process-specific customization andevaluate tools based on business outcomes rather than software benchmarks. They expect systems that integrate with existing processes and improve over time. Vendors meetingthese expectations are securing multi-million-dollar deployments within months. While most implementations don't drive headcount reduction, organizations that havecrossed the GenAI Divide are beginning to see selective workforce impacts in customersupport, software engineering, and administrative functions.In addition, the highest-performing organizations report measurable savings from reduced BPO spending andexternal agency use, particularly in back-office operations. Others cite improved customerretention and sales conversion through automated outreach and intelligentfollow-upsystems. These early results suggest that learning-capable systems, when targeted atspecific processes, can deliver real value,even without major organizational restructuring. 3THEWRONGSIDE OF THEGENAIDIVIDE:HIGHADOPTION,LOWTRANSFORMATION Takeaway:Most organizations fall on the wrong side of the GenAI Divide,adoption is high,but disruption is low. Seven of nine sectors show little structural change. Enterprises arepiloting GenAI tools, but very few reach deployment. Generic tools like ChatGPT are widelyused, but custom solutions stall due to integration complexity and lack of fit with existingworkflows. The GenAI Divide is most visible when examining industry-level transformation patterns.Despite high-profile investment and widespread pilot activity, only a small fraction oforganizations have moved beyond experimentation to achieve meaningful businesstransformation. 3.1THEDISRUPTIONREALITYBEHIND THEDIVIDE Takeaway:The GenAI Divide manifests clearly at the industry level,despite GenAI'svisibility,