무슨 일이 일어났나요?
ZDNET reported on Deloitte research based on a survey of 501 senior business leaders involved in their organizations’ AI strategies or implementations. According to ZDNET, 42% of organizations were testing small numbers of agents, 43% were expanding deployments across functions, and 15% had reached scaled, orchestrated multi-agent adoption across areas including customer service, IT and engineering. The supplied source does not include Deloitte’s primary report, so these figures have not been independently confirmed here.
ZDNET published the report on Aug. 24, 2026, describing what it called the latest Deloitte research on agentic AI adoption in U.S.-based organizations. The article said the research surveyed 501 senior business leaders who were involved in driving AI strategies or implementations. According to ZDNET, 42% of organizations were testing small numbers of agents, 43% were expanding deployments across functions, and 15% had achieved scaled, orchestrated multi-agent deployments across customer service, IT and engineering. The source presents these figures as survey findings, not as an independently audited census of all U.S. businesses. Deloitte’s underlying report, questionnaire and respondent breakdown are not included in the supplied material, so AI Understanding has not independently confirmed the figures or methodology.
ZDNET reported that nearly two-thirds of the leaders surveyed were reevaluating their business models because of advances in agentic AI, while half had a clear view of a future operating model powered by agents. The article identified three leading obstacles to scaling: a lack of a unified and accessible data foundation, cited by 72% of respondents; difficulty trusting and governing agents, cited by 70%; and the cost and complexity of integration, cited by 67%. ZDNET also said that only a minority of organizations considered key foundations ready for agentic AI, including vision and strategy at 36%, technology infrastructure at 34%, data foundations at 32%, and risk, security and governance at 26%.
The article further reported that only 16% of business leaders said their current processes were prepared for agentic adoption. ZDNET said 74% expected nearly half of business processes to be redesigned or rebuilt around AI agents by 2030, while 61% believed most processes would be powered by agents and that many of those agents would operate with little or no human involvement. It also reported that only 31% expected to redesign and rebuild processes around agents by 2028. On workforce preparation, ZDNET said half of leaders believed their organizations were not investing enough in workforce transformation, 43% anticipated major job disruption, 71% were working on baseline agent literacy and 65% were pursuing upskilling or reskilling. These are reported expectations and self-assessments; the source provides no independent measurement of actual deployments, job losses or productivity outcomes.
왜 중요한가요?
The report frames agentic AI adoption as an organizational redesign problem, not simply a software deployment. If the survey is representative, companies moving from experiments to production will need stronger data foundations, governance, integration plans, redesigned workflows and sustained investment in employee training. It also suggests that many executives anticipate substantial changes to job roles and operating models, while the evidence supplied does not establish whether those expectations will be realized.
The central significance of the report is its emphasis on organizational capacity. Agentic systems can perform sequences of tasks, use tools and interact with business systems, so deployment can affect how work is assigned, reviewed and approved. A company may be able to connect an agent to an existing workflow without having reliable data, clear permissions or a practical way to intervene when the system fails. The barriers identified by ZDNET—data access, governance and integration—therefore concern the conditions under which an agent can be trusted with consequential work, not merely whether a company has purchased an AI product.
The workforce findings are also consequential because they describe adoption as a change in job design rather than a narrow automation project. If routine and structured tasks become more autonomous, employees may spend more time checking outputs, handling exceptions, managing customer relationships or defining goals. That transition could create new responsibilities, but it could also reduce demand for some tasks or roles. ZDNET reported that 43% of surveyed leaders expected major job disruption, yet the article supplied no details about which occupations were included, how disruption was defined or whether the expectation referred to job elimination, changed duties or redeployment.
The projections through 2030 should be treated as planning signals rather than forecasts established by observed results. A survey can show what leaders believe and what organizations say they are preparing to do, but it cannot by itself demonstrate that agents will achieve reliable autonomy, deliver savings or redesign most business processes. The source also does not report failure rates, security incidents, accuracy measures, customer effects, spending or comparative results between agent-based and conventional workflows. Those omissions matter because the business case for deployment depends on performance in real operating environments, including exception handling and human oversight.
대화형 메커니즘: 실제로 작동하는 방식
이 개발의 이면에 있는 기본 기술을 대화식으로 살펴보세요.
crm_get_transaction(id='4092').An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
다음에 무엇을 볼 것인가
The key signals will be independently verifiable deployments, measurable operational outcomes, changes to workforce responsibilities and evidence that companies are redesigning processes rather than placing agents on top of unsuitable legacy systems. Organizations will also need to disclose how they control autonomous actions, measure errors, manage data access and account for model and costs. The source does not identify the surveyed companies, provide the full methodology or show independent validation of the projections through 2030.
The first priority is independent verification of the Deloitte findings. Useful follow-up would include the original report, the survey instrument, field dates, respondent industries, company sizes, geographic scope and definitions for terms such as “scaled,” “orchestrated” and “multi-agent.” Without those details, the percentages are informative but difficult to compare with other adoption surveys or generalize to the broader business population.
Companies’ process redesign decisions will be another important signal. ZDNET reported that organizations often layer agents onto existing processes even though redesign may be necessary for greater autonomy. Watch for public case studies that describe the original workflow, the agent’s permissions, human checkpoints, error handling, data sources, deployment scale and measured outcomes. Claims of transformation will be more meaningful when they include concrete evidence such as completion rates, review workloads, time to resolution, error costs and changes in staffing or responsibilities.
Governance and workforce effects should be monitored together. Organizations adopting agents will need to explain who can authorize actions, how access to sensitive data is limited, how activity is logged, how models are evaluated after updates and how people can reverse or contest automated decisions. They should also report what training employees receive and whether workers are redeployed, have their duties expanded or face reductions. The source gives no evidence yet that the reported expectations for widespread autonomous processes by 2030 will occur, so practical results—not executive projections alone—will determine the significance of this trend.