AI价值缺口：为何AI部署速度超过价值实现——以及组织如何缩小差距

The AI value gap Why AI deployment is outpacing value realization –and how organizations can close the gap Companies across industries are facing a widening AI value gap: They areinvesting heavily in the technology, but financial returns are not keeping pace.This dynamic resembles a form of "profitless prosperity" in AI. Almost 90 percent To understand what lies behind this value gap, Roland Berger's AI Lab surveyedmore than 200 top executives in December 2025 across five major industries andfive geographies, focusing on the adoption of generative AI applications andagentic AI systems. The results show that companies fall into four distinct groups, In this study, we diagnose the root causes of today's profitless prosperity andshow what organizations can do to close the AI value gap. Performance varies bygeography and industry, with Japan leading globally, the DACH region (Germany, Contents The state of AIadoption Widespread deployment, rising budgets,limited financial returns In just a few short years, organizations across industrieshave moved from exploring the possibilities of AI to production. The shift from chat-based generation to action-oriented execution is underway, and more than half oforganizations have now deployed autonomous agents in a once an innovation project at the edge of the enterpriseis increasingly becoming a standard operating condition,with companies scaling both generative AI (GenAI) andagentic AI into real business environments. Yet this surgein activity masks a growing divergence between successfuland less successful firms: While AI momentum is now Adoption, however, is only part of the story. This strategicmomentum is also being matched by a surge in investment.Budgets are rising quickly, with over a quarter of firmsinvesting more than USD 5 million in GenAI and agenticAI in fiscal 2025. That share increases to 34 percentin companies' plans for fiscal year 2026, reflecting a This study examines both generative AI and agentic AIadoption. Organizations increasingly tackle both as part ofasingle AI transformation agenda. Yet agentic systems,which autonomously plan, act and adapt, represent the A common narrative is that the market is engaged in asimple "build or buy" debate. In reality, the dominantimplementationmodel is neither pure in-housedevelopment nor full outsourcing, but what we might call"hybrid orchestration". Chosen by almost 40 percent ofrespondents, off-the-shelf solutions remain the mostcommon approach, closely followed by hybrid models that To understand how far these developments haveprogressed, we surveyed 203 top executives across fivemajor industries and five geographies in December 2025.1The results show that executive engagement with AI is nownearly universal: 99 percent of firms in our survey reportformal leadership involvement in AI initiatives, while98 percent have already defined a strategic investment A Spending on AI is surging AI budget allocation trajectory by investment tier(2025 vs. 2026, % of all companies) significantly longer for returns. Over time, this velocity gapcompounds into a breakeven trap. Nearly three-quarters ofthe surveyed firms report that their AI projects face delaysor extended timelines, while just 14 percent consistently FAILURE TO DELIVER Despite this decisive shift in the market, companies are oftendisappointed with the results. The survey data reveals whatwe call a "velocity gap": Companies are bringing AI intoproduction faster than they are capturing value from it. Thus,while 37 percent of organizations reach production within The pattern becomes even clearer when firms compareoutcomes against their initial projections. What emerges B The efficiency-to-value conversion gap Operational efficiency vs. financial returns (% of all companies) is an efficiency-to-value conversion gap: Organizations arefar better at achieving operational improvements thanconverting those gains into financial impact. The dominant,buy-heavy approach can deliver process optimization, but why, we must examine how the market has splintered intodistinct realities, with some players trapped in this patternand others already breaking through. This is the subject of In short, companies have learned to make AI work, but theyhave not yet learned how to make it pay off. To understand Unevenoutcomes Four AI performance archetypes defined bystrategy and execution W into four distinct segments based on their strategic intentand execution capability. Each segment reflects a differentrelationship between investment and returns – and each, success. Although AI commitment is now widespread,outcomes are increasingly uneven. To make sense of thisperformance split, we grouped the 203 firms in our survey C High investment does not guarantee high returns At the top right are theIndustrializers,representing aroundten percent of the sample. They set the performancebenchmark for large-scale AI deployment, havingsuccessfully synchronized speed-to-pr