代理型AI的崛起：信任何以成为人机协作之关键

Capgemini Research Institute 2025 Table ofcontents majority of survey respondents (900 out of 1500). This exerciseshowed results similar to the original data, confirming strongadoption numbers. In the near term, AI agents are expected tosee most extensive adoption in customer service, IT, and sales,expanding into operations, R&D, and marketing over the nextthree years. But expectations for highly autonomous systemsremain limited. Our survey shows that, in 12 months, only 15%of all business processes are expected to operate at Level 3(semi-autonomous) to Level 5 (fully autonomous). This shareis expected to grow to 25% by 2028. Most AI agents currentlyoperate at low levels of autonomy, primarily as simple agentsor semi-autonomous agents. Notably, AI agents can delivertangible value even at intermediate levels of autonomy.Moreover, trust in AI agents is declining.Only 27% oforganizations express trust in fully autonomous AI agents,from 43% 12 months ago. This is potentially a reflection ofbusiness reality taking hold after the initial enthusiasm andoverconfidence in agentic AI capabilities. Ethical concernsaround AI, such as data privacy, algorithmic bias, and the“AI black box,” are prevalent, but few organizations actCapgemini Research Institute 2025 ExecutivesummaryAI agents constitute one of the fastest-emerging technologicaltrends (see page six,“What are AI agents and what is agenticAI?”). Unlike traditional AI applications, which often act astools focusing on narrow, predefined tasks, agentic AI canmanage and execute end-to-end processes with less humanintervention, evolving from tools to team members andmarking a new era in enterprise productivity, efficiency, andgrowth.Based on our research findings and current adoptiontrajectories, we project that AI agents could generate up to$450 billion in economic valuethrough revenue uplift andcost savings across surveyed countries by 2028. Competitivemomentum is clearly building: 93% of leaders believe thatthose who successfully scale AI agents in the next 12 monthswill gain an edge over industry peers.Already, 14% of organizations have implemented AI agentsat partial (12%) or full scale (2%) and nearly one-quarter(23%) have launched pilots, while another 61% are preparingfor or exploring deployment. As these numbers showed aremarkably fast adoption, we reconfirmed the responses of a Executivesummarydecisively to mitigate them. Furthermore, only half claimsufficient knowledge of AI agent capabilities, and evenfewer can pinpoint where agents outperform traditional AIor automation. Equally worryingly, fewer than one in fiveorganizations report high levels of data-readiness, and over80% lack mature AI infrastructure, significantly limiting theirability to scale agentic systems effectively.Nonetheless, in 3 years, organizations expect to have AI agentsas members within human-supervised teams.This means thatAI agents must be seen not as tools, but as part of the team.Such human-AI agent collaboration enhances the value of AIagents, driving transparency, trust, and positive outcomes. But61% of organizations report rising employee anxiety about theimpact of AI agents on their employment prospects, and overhalf believe AI agents will displace more jobs than they create.Despite this, fewer organizations are prioritizing reskilling orworkforce restructuring. To harness the full potential of AI agents, organizations mustmove beyond the hype and work toward:•Redesigning processes to deploy AI agents effectively andreimagining business models•Transforming the workforce and organizational structure tofully onboard a new, agentic workforce as team members•Striking the right balance between agent autonomy and humaninvolvement•Strengthening data and technological foundations to scale AIagents•Ensuring AI agents operate within defined scope ofexecution,and remain traceable and explainable to earn trust•Developing and integrating ethical AI that address the risksposed by autonomous AI agents.Capgemini Research Institute 2025 We’d also liketo thank themany industryexecutives whoshared theirvaluable insightswith us. Lynn CompHead of Global Sales and GTM(AI Center of Excellence),IntelJoji PhilipDirector AI/ML Products,EricssonCapgemini Research Institute 2025 Jason GelmanDirector of productmanagement, Vertex AI,Google CloudItai AsseoHead of Incubation and BrandStrategy (AI Research),Salesforce Anna KoppDigital Lead Germany,MicrosoftEric PaceHead of AI,Cox Communication Dr. Suraj Srinivasan,Philip J. Stomberg Professor ofBusiness Administration and Chairof the MBA Elective Curriculum,Harvard Business SchoolPreetha SekharanVP, Digital Incubator(Applied AI andTransformation),Unum Susan EmersonSenior VP, AI Product,SalesforceDaniel VassilevCo-Founder and Co-CEO,Relevance AI Dr. Walter SunSVP, Global Head of AI,SAPVishal SinghviDirector, StrategicInitiatives – Gen AI,Microsoft Dorit ZilbershotGVP, AI Experiences & Innovation,ServiceNow Who shouldread this reportand why?This r