Now, that conversation is changing. As AI budgets continue to rise and deployments become more widespread, enterprises are facing a more difficult question: What returns are those investments actually generating?
The question comes at a time when AI adoption is accelerating across industries. According to Nvidia’s State of AI report published in May this year, 64 per cent of enterprises are actively using AI across their operations. Yet widespread adoption has not automatically translated into business transformation. According to the Nvidia’s report, organisations struggle with scaling AI projects beyond pilot stages, measuring return on investment (ROI), and redesigning workflows to take full advantage of the technology.
A similar trend is captured by Deloitte’s in its State of AI in the Enterprise: The Untapped Edge report, which it published this year in January.
Together, the findings suggest that enterprise AI is entering a new phase — one where success will be measured less by deployment numbers and more by tangible business outcomes.
What do the latest surveys reveal about enterprise AI adoption?
One of Deloitte’s key findings is the rapid expansion of AI access across the workforce. In just one year, the proportion of employees with approved access to AI tools increased from less than 40 per cent to around 60 per cent. Among the most advanced organisations, more than 80 per cent of employees now have access to AI technologies.
Nvidia’s findings point in a similar direction. Based on responses from organisations across sectors, including healthcare, retail, manufacturing, telecommunications, and financial services:
-
64 per cent said they are already using AI in their operations. -
28 per cent remain in pilot or evaluation stages. -
Active AI adoption stands at 63 per cent in the Asia-Pacific region, including India.
However, providing access does not automatically translate into regular use. Deloitte notes that fewer than 60 per cent of employees who have access to AI tools actually use them in their daily work. This figure has remained largely unchanged from the previous year, suggesting that adoption challenges extend beyond technology availability.
Larger organisations appear to be further ahead. According to Nvidia, 76 per cent of companies with more than 1,000 employees actively use AI. These firms also report stronger returns, largely because they have greater resources to invest in infrastructure and can dedicate leadership attention to scaling projects beyond the pilot stage.
Why are organisations struggling to move AI beyond pilot projects?
While AI experimentation is widespread, turning successful pilots into large-scale deployments remains difficult.
In Deloitte’s survey, only 25 per cent said their organisations have successfully moved at least 40 per cent of AI experiments into production. However, 54 per cent expect to reach that milestone within the next three to six months, suggesting confidence that deployment efforts are accelerating.
The challenge lies in the difference between an AI pilot and a production-scale deployment. Pilot projects can often be developed by small teams using controlled datasets and limited infrastructure. Scaling those projects across an enterprise is far more complex, according to the report.
Production deployment requires integration with existing systems, security and compliance reviews, monitoring mechanisms, and ongoing maintenance. Solutions that perform well in testing environments may encounter unexpected issues when exposed to real-world conditions, larger datasets, and higher user volumes.
As a result, initiatives that appear successful in the pilot stage can face significant hurdles when moving into full-scale operations.
This has led to what some industry leaders describe as “pilot fatigue”, where organisations continuously test new AI ideas without a clear roadmap for scaling them into business operations.
Is AI generating revenue or mainly improving productivity?
Both reports suggest that AI is already delivering measurable business benefits, particularly through productivity improvements and cost efficiencies. However, the gap between operational gains and revenue generation remains significant.
While many organisations are seeing AI improve existing processes, far fewer have translated those gains into meaningful business transformation or new revenue streams.
Nvidia’s research suggests that AI is moving beyond experimentation, with organisations reporting gains in revenue, cost efficiency, and employee productivity. However, significant financial returns remain limited to a smaller group of companies.
NVIDIA’s State of AI report
Deloitte’s findings present a more measured picture. While two-thirds of organisations report efficiency and productivity gains from AI, only 20 per cent say they are generating new revenue through AI today, even though nearly three-quarters expect to do so in the future.
The findings suggest that AI is currently being used more as a tool for optimisation than as a catalyst for new business models, with only about one-third of organisations fundamentally transforming products, services, or operations through AI.
Why is agentic AI emerging as the next enterprise battleground?
Another major trend highlighted by both studies is the rise of agentic AI — systems capable of independently carrying out multi-step tasks and making decisions within defined boundaries.
Deloitte found that 23 per cent of organisations already use agentic AI to some extent. Within two years, that figure is expected to rise to 74 per cent.
Nvidia reported particularly strong adoption in the telecommunications and retail sectors.
The appeal of agentic AI lies in its ability to automate complex workflows rather than simply assist employees with individual tasks.
In healthcare, for example, Nvidia highlighted a medical-assistant AI that significantly reduced documentation errors while lowering clinician workload.
Yet governance remains a major concern. Only 21 per cent of Deloitte respondents said they have mature frameworks for managing autonomous AI systems.
Why do talent and workforce challenges remain major barriers?
Technology alone is not the biggest challenge. Both reports identify people and skills as critical barriers to AI success.
Deloitte found that a lack of workforce skills is the most commonly cited obstacle to AI integration. Despite this, fewer than half of organisations are making significant changes to talent strategies. Most are focused on improving AI literacy rather than redesigning jobs and workflows around AI capabilities.
The findings reveal a significant disconnect. While 82 per cent of organisations expect at least 10 per cent of jobs to become fully automated within three years, 84 per cent have not yet redesigned roles to reflect that reality.
Entry-level positions, often used as stepping stones for career development, are among those most likely to be automated.
Nvidia’s survey echoes these concerns. Data-quality issues, shortages of AI expertise, and uncertainty around ROI remain among the most frequently cited barriers to scaling AI initiatives.
According to S&P Global’s The AI and Labor Landscape 2026 report, companies are primarily adopting AI to boost productivity rather than deliberately reduce headcount. Yet the near-term impact on employment remains mixed.
Globally, the balance of firms reporting AI-related job losses over the past year was five percentage points higher than those reporting job gains.
At the same time, many organisations continue to view employee productivity, process efficiency, and cost reduction as the primary measures of AI success.
This suggests that AI is beginning to reshape how organisations structure work and evaluate performance.
Why are enterprises continuing to increase AI spending?
Despite the obstacles, enterprise confidence in AI remains strong. Deloitte found that 84 per cent of organisations plan to increase AI investments, while Nvidia reported that 86 per cent expect AI budgets to grow in 2026. Nearly 40 per cent anticipate budget increases of at least 10 per cent.
A notable theme emerging from Deloitte’s research is the growing importance of sovereign AI. More than 80 per cent of respondents view it as strategically important.
What does the next phase of enterprise AI look like?
Taken together, the findings from Deloitte, Nvidia, and S&P Global suggest that enterprise AI is entering a new phase. The debate is no longer about whether organisations should adopt AI. With investments rising and adoption becoming widespread, the focus is shifting to whether companies can generate measurable returns from those investments.
Productivity gains and cost savings are becoming increasingly visible, but revenue growth, workforce transformation, and large-scale deployment remain works in progress.
At the same time, the findings make it clear that technology alone will not determine success. Organisations must redesign workflows, develop AI skills, establish governance frameworks, and rethink how work gets done.
The next phase of AI is less about deploying new tools and more about organisational transformation. The companies that succeed will be those that can move beyond experimentation and convert AI investments into sustained business outcomes.
