2025年代理型AI报告：为智能体时代做好准备（英）

the Agentic Era Authored by Thomas H. Davenport and Ian Barkin uipath.com Executive Summary The promise of enterprise AI—and particularly generative AI (GenAI)—has reachedan inflection point. While demand for this technology has skyrocketed in the lastfew years, the applications for and value of GenAI as a standalone technology mayhave plateaued. Large language models (LLMs) can be useful for content creationbut are limited in their ability to provide actionable intelligence to workers,integrate with other business applications and systems, and ensure governance, Enter the agentic era, where AI agents, automation, and people combine toperform more complex autonomous business workflows.But how will this newage of AI transform the enterprise? And are IT and business workforces prepared? is underway—though thereare risks that enterprises mustconsider to harness the full senior technology leadersregarding their awarenessof and interest in agentic AI.The following report indicatesexcitement for agentic AI is Key Findings The top perceived benefits of agentic AI include: The top concerns of agentic AI include: The top limitation of existing AItools is a lack of integration with 2025UiPathAgentic AI The state of AIand automation today Before diving into the stateand future of agentic AI, it’simportant to understand theextent to which AI and AI-adjacent technologies areWhich of the following AI or AI-adjacent technologies does Almost all respondents (92%)said it is completely or verylikely that their organizationswill increase investments in AIor AI-adjacent technologiesover the next year. IT processautomation ranked as the top 2025UiPathAgentic AI Concerns with existing AI While most IT leaders have seenvalue from their automationand AI deployments, they havealso experienced challenges—namely, security, development Their perceptions of thegreatest limitations of currentAI technologies include a lack ofintegration with other businessapplications (61%); an inabilityof AI tools to learn and adaptwithout human intervention With GenAI and LLM solutionsspecifically, respondents saidthey were most concerned withdata quality (47%); IT securityrisk (33%); lack of explainabilityof results (31%); and the relatedconcern of lack of trust in results(30%). 2025UiPathAgentic AI Why technology leadersare turning to agentic AI agentic AI empowers a newbreed of autonomous agentsbestowed with cognitiveabilities such as learning,problem solving, and decision To address the concerns andchallenges with traditional AI,many companies are exploringagentic AI—a groundbreakingadvancement in artificial decisions and taking actions,enabling more intelligentautomation, and performingmore tasks with little or no automating tasks and complexoperations. Technologies likerobotic process automation(RPA), LLMs, and predictiveanalytics are complementary,and will be combined into For our purposes,the key components of agentic AI include the abilities to: Act autonomouslywith little or no human Use AI skills to plan andcarry out complex tasksand make decisions The survey shows that inmany organizations, agenticAI is here today: 37% ofrespondents say they arealready using the technology, Network with andorchestrate other agents andsystems to enable dynamicand complex processes Almost half of IT executives(45%) say they are ready toinvest in agentic AI in 2025,with 32% predicting they willlikely do so within six months Perform a clearlyspecified task withspeed and quality Fiserv, the global fintechand payments company, hasalready developed an earlyagentic use case.SharbelShaaya, the company’sDirector of AI Operationsand Intelligent Automation,described an applicationthat can look up and assignmerchant codes to companies interprets the research, and athird component assigns themerchant code. More than99% of the actions require nohuman validation, which is aconsiderable improvementover human-only workflows.But Shaaya cautions that thecompany had to do all thehard work of creating andtuning prompts, structuring The strong interest in agenticis not surprising given thesetypes of early results—andthe anticipated value of thetechnology. When askedwhich benefits of agentic AIwould be most appealingto their businesses, surveyrespondents named IT leaders are alsoreceiving a push to fast-track deployments: 84% ofrespondents indicate thatthey feel somewhat, very,or extremely pressured to Similarly, they found themost appealing capabilitiesof agentic AI to be increasedautomation (55%), improvedproblem solving (53%), andimproved accuracy andreduced errors (53%). Ninetypercent of respondentsalso say they have business One veteran automation andAI leader is bullish on agentic. Andy Fanning, who until recentlyled GenAI, intelligent automation,business transformation, andtechnology R&D for Cigna andis nowleading an AI startup,Optura.ai, views agentic systems asa natural progression in computing.He explains that AI agents represent