人工智能现状：组织如何重组以获取价值
How organizations are rewiring to capture value
Alex SinglaAlexander SukharevskyLareina Yee
March 2025
Organizations are beginning to create thestructures and processes that lead tomeaningful value from gen AI. Whilestill in early days, companies are
Organizations are startingto make
organizational changesdesigned togenerate future value from gen AI, and
that organizations are beginning to take steps that drivebottom-line impact—for example, redesigning workflows asthey deploy gen AI and putting senior leaders in critical roles,such as overseeing AI governance. The findings also showthat organizations are working to mitigate a growing set ofgen-AI-related risks and are hiring for new AI-related roleswhile they retrain employees to participate in AI deployment.Companies with at least $500 million in annual revenue
How companies are organizingtheir gen AI deployment—
Our survey analyses show that a CEO’s oversight of AI governance—that is, the policies,processes, and technology necessary to develop and deploy AI systems responsibly—is oneelement most correlated with higher self-reported bottom-line impact from an organization’sgen AI use.1That’s particularly true at larger companies, where CEO oversight is the element withthe most impact on EBIT attributable to gen AI. Twenty-eight percent of respondents whose
The value of AI comes from rewiring how companies run, and the latest survey shows that, outof 25 attributes tested for organizations of all sizes, the redesign of workflows has the biggesteffect on an organization’s ability to see EBIT impact from its use of gen AI. Organizations arebeginning to reshape their workflows as they deploy gen AI. Twenty-one percent of respondents
Twenty-eight percentof respondentswhose organizations use AI reportthat their CEO is responsible for
Twenty-one percentofrespondents reporting gen AI use
by their organizations say theirorganizations have fundamentally
McKinsey commentary
Alexander Sukharevsky
Senior partner and global coleader of QuantumBlack, AI by McKinsey
The more we seeorganizations using AI, the more we recognize that it takes a top-down process to really move the needle. Effective AI implementation starts with a fullycommitted C-suite and, ideally, an engaged board. Many companies’ instinct is to delegate
There are several reasons for this. The first is that getting real value out of AI requirestransformation, not just new technology. It’s a question of successful change managementand mobilization, which is why C-suite leadership is essential. It’s also a potentially expensivetransformation, requiring intensive use of sometimes scarce resources and talent. A lot
As organizations become more fluent with AI, it will essentially become embedded in allfunctions, leaving leadership to focus on higher-level tasks like impact monitoring and talent
Organizations are selectively centralizing elements of their
The survey findings also shed light on how organizations are structuring their AI deploymentefforts. Some essential elements for deploying AI tend to be fully or partially centralized(Exhibit 1). For risk and compliance, as well as data governance, organizations often use a fullycentralized model such as a center of excellence. For tech talent and adoption of AI solutions,on the other hand, respondents most often report using a hybrid or partially centralized model,
Web <2024>
Exhibit 1Exhibit <1> of <14>
Risk and data governance are two of the most centralized elements ofdeploying AI solutions, whereas tech talent is often hybrid.
McKinsey & Company
Twenty-seven percentofrespondents say employeesat their organizations review allcontent created by gen AI beforeit is used, and a similar share says
Organizations vary widely in how they monitor gen AI outputs
Organizations have employees overseeing the quality of gen AI outputs, though the extentof that oversight varies widely. Twenty-seven percent of respondents whose organizationsuse gen AI say that employees review all content created by gen AI before it is used—forexample, before a customer sees a chatbot’s response or before an AI-generated image isused in marketing materials (Exhibit 2). A similar share says that 20 percent or less of gen-
¹Question was asked only of respondents whose organizations use AI in at least 1 function, n = 1,229. Figures were calculated after removing the share who said“don’t know/not applicable.”AI-produced content is checked before use. Respondents working in business, legal, andother professional services are much more likely than those in other industries to say that all
Exhibit 2Web <2024>
