From Pilots to Payoffs: How AI is really delivering value in 2025
Roughly a quarter of survey respondents say their company is scaling at least one agentic system - systems that can plan multi-step tasks and act in the real world. Results of the survey on the State of AI in 2025 from McKinsey find the AI landscape appears to be in its infancy: many organisations are using AI, but few are seeing any significant financial benefits at this stage. The research was carried out in 2025, obtaining survey responses from 1,993 participants in 105 nations over a range of organisation types.
Legal risks
The legal risks associated with agentic AI systems are numerous. They include challenges related to data protection, biased behaviours, supply chain dependency, restricted human control, higher risk for conducting transactions, transparency challenges, and more.
Adoption and impact
The report highlights that, although interest in AI agents is high, adoption is still in the early stages and is typically confined to one or two functions. Despite nearly nine in ten organisations reporting regular AI use in at least one function, only around a third are scaling AI across the enterprise. Fully scaled cases are still rare. Agentic AI is moving from buzz to trials. Of the quarter of respondents who say their company is scaling at least one agentic system, it is most often in IT and knowledge management. The technology, media, telecoms and healthcare sectors are leading the way in this respect.
Value is appearing at the use case level. Cost reductions are most commonly reported in software engineering, manufacturing, and IT. Meanwhile, revenue uplift is more often seen in marketing and sales, strategy and corporate finance, and product development. Only 39% report any enterprise level impact on earnings, and for most of those, the effect is modest.
What high performers do differently
A small cohort report significant value and an impact of more than a five percent on enterprise level earnings from AI. These high performers set bold transformation goals that go beyond efficiency. They redesign workflows to embed AI, and scale faster across a wider range of functions. Over a third of them spend more than 20% of their digital budget on AI. High performers combine strong leadership and ownership with agile delivery and robust technological and data and foundations, as well as clear human validation controls.
Risks and workforce signals
The report reveals that opinions differ on the impact of AI on enterprise headcount. Larger organisations expect more change in the next year than was observed in the last year, with about a third anticipating reductions, while a meaningful minority expect increases. Reported negative consequences are common, led by inaccuracy. Risk mitigation efforts are rising, notably on privacy, cybersecurity, compliance, and AI inaccuracy. Given their ambition, it is unsurprising that high performers encounter more risks, yet they also mitigate more.
The centre of gravity is shifting from experimenting with generative AI to building durable, scaled capabilities. Organisations that unlock value treat AI as a catalyst for growth and innovation, rather than just as a cost-cutting tool. They do the unglamorous work of redesigning workflows, enabling human oversight, and investing in platforms and talent.
