Research — Oct 27, 2025
Generative AI shows rapid growth but yields mixed results
This report examines the current state of generative AI, highlighting accelerated expansion and the diverse results organizations have experienced. It is the first of two reports investigating where organizations have deployed AI over the past 12 months, the performance of those projects, and plans for the next year. To learn more or to request a demo, visit spglobal.com/451Research.
Introduction
The generative AI landscape has undergone a significant transition over the past year. If 2024 was characterized by organizations experimenting with the technology, in 2025 we then see companies both more deeply integrating these technologies and expecting tangible results. This report is the first of two that investigate the AI strategies organizations are adopting and their reflections over the past 12 months. It places a spotlight on the varied outcomes from the initial wave of generative AI investments and the challenges many organizations are facing. The second report, “The next chapter in generative AI,” will delve into what leading organizations are doing differently and highlight the technological trends poised to enhance the impact of generative AI.
About this report
Reports such as this showcase insights derived from a variety of market-level research inputs, including financial data, M&A information, and other market data sources both proprietary to S&P Global and publicly available. This input is combined with ongoing observation of markets and regular interaction with vendors and other key market players.
This report specifically includes data from the following sources. See the Methodology section at the end of the report for more details.
- Voice of the Enterprise: AI & Machine Learning, Use Cases 2025 — This online survey of 1,006 respondents was conducted Oct. 21–Nov. 25, 2024.
- Voice of the Enterprise: AI & Machine Learning, Use Cases 2024 — This online survey of 1,001 respondents was conducted from Nov. 14–Dec. 14, 2023.
- Voice of the Enterprise: Workforce Productivity & Collaboration, Employee Engagement 2024 — This online survey of 589 respondents was conducted from Oct. 17–Nov. 12, 2024.
Key findings
- Generative AI has rapidly proliferated, surpassing even last year’s ambitious forecasts. As of late 2024, 60% of organizations that are investing in AI have implemented generative AI, outpacing longer-standing AI categories such as rules- based (54%) and pattern-recognition (51%) models.
- The accelerated uptake of generative AI is driving a significant uptick in project failure rates. The proportion of companies abandoning most of their AI initiatives before they reach production has risen from 17% to 42%, with the average organization scrapping 46% of their proof-of-concept projects prior to production.
- Generative AI is delivering value, but its impact is uneven. While 46% of companies report no single enterprise objective has produced a “strong positive impact” from generative AI initiatives, 19% report strong positive impact across most objectives.
- Significant obstacles to successful generative AI investments include cost, privacy concerns and trust issues. Budget limitations and confidence in accuracy are leading issues, with 29% of organizations identifying these as significant challenges.
The Take
In 2024, a clear message about generative AI emerged across the vendor ecosystem. That message characterized generative AI as a transformative technology that organizations could use to drive greater efficiencies and gain competitive advantage. The narrative divided organizations into “haves” and “have-nots” — and senior executives strove to be in the former camp. Ample budgets and executive support facilitated a rapid transition from experimentation with tools to broad availability across teams, with some organizations even exposing capabilities to customers.
As we enter 2025, many organizations are reflecting on their early investments and finding that the results have not met their lofty expectations. While the technology is delivering value, how that value measures up against costs can be unclear. From the “haves” and “have-nots,” we see a new set of camps emerging: the “cans” (organizations effectively leveraging the technology) and “cannots” (organizations struggling to do so).
The second report in our series focuses on the “cans” and trends that may drive greater future value. This report emphasizes the “cannots,” identifying skills shortages, a lack of targeted application, user rejection and poor accuracy as core challenges.
Generative AI experiences explosive growth, outstripping ambitious forecasts
Generative AI has captured the collective imagination since the November 2022 debut of OpenAI’s ChatGPT, the first of a set of popular generative AI chatbots that demonstrated significant accessibility and creativity, and consequently, broad
uptake. It was soon followed by a wave of popular image generator applications. These developments prompted senior executives to explore applications of this general- purpose technology in their organizations’ digital roadmaps.
This exploration rapidly transitioned into formal adoption. As we enter 2025, most organizations that are investing in AI have generative AI in production. Of this group, 27% report organization-wide adoption, while 33% say it is limited to specific departments or projects. As Figure 1 illustrates, this represents accelerated uptake compared with the previous year, when the figures were 13% and 28%, respectively.
The plurality of respondents expects generative AI to be fully integrated across their organizations by Q4 2025.
Figure 1: Generative AI shows accelerated uptake
Q. Please select the appropriate level of usage for each AI type.
Q. What is the expected status of each AI type in your organization 12 months from now?
Q. Which of these statements most accurately reflects the use of generative AI at your organization? Base: All respondents.
Source: 451 Research’s Voice of the Enterprise: AI & Machine Learning, Use Cases 2024 and 2025.
The adoption rate of generative AI has swiftly surpassed longer-standing forms of AI. While 60% of respondents report implementing generative AI, this figure stands at 54% for rules-based models and 51% for pattern-recognition models. Respondents from the software, IT and computer services, and retail sectors show particularly rapid generative AI uptake, with more than one-third of organizations reporting full integration across their organization.
“There has been a significant strategic shift this year to go headfirst into AI. A new AI team was created this year to focus on learning about AI, figuring out how to leverage it effectively, where are the opportunities within our business and in our revenue- generating areas to introduce AI capabilities.”
-IT/engineering manager/staff, 250-499 employees, $1B-$2.49B revenue, financial services
Source: 451 Research’s Voice of the Enterprise in-depth interview, September 2024.
Notably, the telecommunications sector is poised for accelerated growth, with an 18-percentage-point increase in those anticipating full organizational integration of generative AI by late 2025 relative to 2024. In the next 12 months, telecoms plan to deliver generative AI capabilities across application areas including synthetic
data generation, process automation, data visualization and coding assistance. The majority of telecoms (51%) anticipate using generative AI for automating or assisting programming tasks, outpacing the survey-wide result of 38%.
Organizations have rolled out generative AI even more rapidly than their ambitious projections suggested a year ago. According to 451 Research’s Voice of the Enterprise: AI & Machine Learning, Use Cases 2024 survey, 47% of organizations anticipated they would deploy generative AI tools by the fourth quarter of 2024. Our latest survey, fielded in the fourth quarter of 2024, reflects a 60% deployment rate — 13 percentage points higher than expected.
Several factors are fueling rapid adoption, related both to opportunity and risk. The two main challenges organizations face in adopting generative AI are data privacy and security risks, each identified by 38% of organizations. In this environment of
pronounced risk awareness, migrating users to tools with greater organizational control is top-of-mind. Another factor is the speed with which existing enterprise software has integrated generative AI, reducing selection and procurement challenges that may have otherwise introduced a bottleneck. According to 451 Research’s Voice of the Enterprise: Workforce Productivity & Collaboration, Employee Engagement 2024 study, just 42% of respondents lacked access to AI assistants or agents. The majority were using some combination of stand-alone tools provided by their employer (26%), unsanctioned tools (19%) and tools integrated in existing software (26%).
“[AI here is] primarily … through third-party product offerings that include generative AI, but we are also working on some homegrown internal applications that make use of large language models … largely applying generative AI to internal products or internal processes.”
-IT/engineering manager/staff, 100,000+ employees, consumer/retail
Source: 451 Research’s Voice of the Enterprise in-depth interview, November 2024.
AI technology is becoming increasingly prevalent in all areas of business. Most organizations investing in AI have implemented the technology in IT operations (63%). Additionally, 45% of organizations have applied AI to customer experience workflows.
Marketing has seen a ramping up of investment. Marketing processes have seen a 7-point increase in the proportion of organizations applying AI year over year. While the most invested-in marketing use case is marketing analysis, which 27% of organizations have applied AI to, marketing campaign management is also an area of focus. Marketing campaigns management is the second most commonly identified use case where organizations have not yet invested in AI, but plan to this year, following financial planning and analysis.
Accelerated uptake derailing many AI projects
Given that organizations are under pressure to deliver generative AI projects at pace, it is perhaps unsurprising to see increasing project failure rates. The proportion of projects failing to graduate from proof-of-concept to production capabilities has increased markedly year over year (see Figure 2). The proportion of companies that abandon most of their AI initiatives has increased from 17% to 42%, with the average organization scrapping 46% of its proof-of-concept projects prior to production.
Figure 2: Surging project failure rates
Q. To the best of your knowledge, what percentage of AI projects in proof of concept at your organization are abandoned before production?
Base: AI/machine learning is in production or proof-of-concept.
Source: 451 Research’s Voice of the Enterprise: AI & Machine Learning, Use Cases 2024 and 2025.
Companies with higher project failure rates are more prone to encountering user resistance. Compared with the overall survey sample, respondents noting an above- average level of project failure are 77% more likely to identify reputational damage, 41% more likely to identify customer resistance, and 36% more likely to identify staff resistance as concerns their organization faces with AI initiatives. This suggests that many organizations have not effectively communicated the value of AI capabilities, or effectively targeted where those capabilities are best applied, hindering their ability to deliver projects effectively. Staff resistance appears particularly problematic in the education and training sector, where 41% of respondents identify it as a concern, compared with the overall figure of 28%.
“There are a significant number of people who are resistant to AI. They don’t use AI to help them in their coding, and they definitely don’t want to help implement AI in the products. I don’t know if it’s a philosophical thing or what, but they seem to be almost religiously opposed to AI for some reason. I’m surprised by that.”
-Senior management, 100-249 employees, software, IT & computer services
Source: 451 Research’s Voice of the Enterprise in-depth interview, November 2024.
“We are constantly pushing the brakes on any AI goals that mostly the C-suite individuals push down the pipe here. The biggest reason why we do that is we don’t really see a real big cost-benefit analysis there, like where the benefits outweigh the costs of integration. And we don’t see too much of a benefit of having AI or purchasing an AI product that everybody would use that gives any benefit over a free product like ChatGPT.”
-IT/engineering manager/staff, $100M-$249.99M revenue, manufacturing
Source: 451 Research’s Voice of the Enterprise in-depth interview, December 2024.
Organizations that experience lower project failure rates tend to consider a broader range of factors when prioritizing projects. While 54% of respondents from organizations with below-average project failure rates consider data availability factors when prioritizing AI use cases, this figure drops to 42% among those with above-average failure rates. There is also a difference of more than 10 percentage
points in the rate of consideration for compliance and risk factors. Organizations with higher-than-average project failure rates consider an average of 4.2 factors in their prioritization process, compared with an average of 4.9 factors for those with better performance. Two factors that do not appear to correlate to project failure rates are time-to-deliver and the presence of existing stakeholder partnerships. This suggests that emphasizing fast delivery may not result in the most effective projects.
Generative AI is delivering value, but impact is uneven
The common narrative influencing enterprise investments in generative AI is that the technology will yield greater value as it becomes integrated into workflows and its user base expands. However, data suggests that this expectation has not been realized. The proportion of organizations reporting a positive impact from generative AI investments has decreased across all assessed enterprise objectives. Although most organizations still recognize a positive impact, the latest survey reveals a significant year-over-
year decline in areas such as revenue growth objectives (76%, down from 81%), cost management (74%, down from 79%) and risk management (70%, down from 74%).
These declines are noteworthy, given the expectation that generative AI initiatives’ benefits would increase with maturation and as organizations transitioned from experimentation to production.
Among respondents whose organizations have invested in generative AI, 46% report that no single enterprise objective has experienced a “strong positive impact.” In comparison, 19% report strong positive impact across the majority of the seven enterprise objectives assessed, which include operational efficiency, customer satisfaction, employee satisfaction, revenue growth, cost management, risk management and sustainability. Considering organizational investments into generative AI, and the clear opportunity costs, the limited performance of these applications for a large proportion of organizations is concerning.
Figure 3: Number of organizational objectives seeing a ‘strong positive impact’ from generative AI
Q. How has generative AI affected your organization’s performance in the following objectives over the past 12 months? Base: GenAI is in use (n=932).
Source: 451 Research’s Voice of the Enterprise: AI & Machine Learning, Use Cases 2025.
Mixed performance also extends to the direct assessment of AI initiatives relative to their set objectives. Companies are measuring significantly more key performance indicators to assess project performance, but projects are not necessarily performing better against them. As an illustration, the proportion of respondents measuring risk reduction, process efficiency, staff satisfaction, risk exposure, model accuracy and achieved social impact KPIs have all increased by more than 6 percentage points
year over year. However, for some of these KPIs, notably including risk exposure, the percentage of companies that consider their projects “very successful” has fallen year over year. Revenue impact, technology uptime, project cost and energy consumption are among the KPIs where organizations report worsening performance.
“A primary pain point right now would probably be implementation of AI in different processes. Some we’ve been really successful in — business automation processes, things like that. But there are others that are just lagging behind.”
-IT/engineering manager/staff, 10-49 employees, $2.50M-$4.99M revenue,software, IT & computer services
Source: 451 Research’s Voice of the Enterprise in-depth interview, November 2024.
The value proposition for generative AI spans many application areas within organizations. Just under half (47%) of organizations using generative AI say it has provided value via data generation in the past 12 months — defined in the survey as creating synthetic data for testing, analysis or training purposes. As Figure 4 illustrates, this investment represents just one facet of the technology’s perceived
value. More than a third of organizations using generative AI say it is creating value in data visualization, process automation, information summarization, content creation, customer interaction, employee training and coding assistance.
The general-purpose and flexible characteristics of the large foundation models supporting generative AI may contribute to the wide range of applications, with respondents identifying value in an average of five areas. In some organizations, a gap between breadth and depth of value might indicate that use cases need better targeting or a sharper focus on more specific application areas.
Figure 4: Areas where generative AI is delivering value
Q. Over the past 12 months, in which of the following areas has generative AI provided meaningful value to your organization, if at all? Please select all that apply.
Base: GenAI is in use (n=950).
Source: 451 Research’s Voice of the Enterprise: AI & Machine Learning, Use Cases 2025.
The diversity of applications is also evident in the generative modalities that organizations are using. The largest proportion of respondents are engaging with text (65%), while images (47%), code (45%) and tabular synthetic data (43%) are also widely used. Other rich media modalities, such as audio and video, are projected to experience rapid growth.
One factor that may be driving the technology’s broad use is the growing number of stakeholders involved in AI project decision-making. As companies improve their
understanding of generative AI, we see a 4-percentage-point year-over-year decline in the proportion of respondents who say machine learning software providers and consultants are involved in AI project approval. However, almost all other stakeholder groups assessed have become more involved over time. This includes internal user groups such as customer service and support, marketing management and sales management. While this can be read as positive — potentially empowering different communities within the business and democratizing access to AI capabilities — a loss of centralized control could result in tool proliferation, increasing costs and a less targeted AI strategy.
Cost, privacy and trust are major bottlenecks to progress
Confidence in accuracy and budget limitations are significant issues that organizations encounter with AI initiatives. The five most common challenges are confidence in accuracy (29%), budget constraints (29%), staff resistance to AI (28%), customer resistance to AI (27%) and skill shortages (27%). Budget limitations have notably risen to the top of the list, up from fourth position last year. This change likely reflects
the increased visibility of software and hardware expenses as organizations move from experimentation to scaled-up AI deployments. Mixed performance may also be prompting a reassessment of costs and benefits of investments.
“I fund more due diligence than we’ve probably ever done on anything is before we turn on GenAI capabilities, just because of the inherent risk of exposing our confidential data.”
-Mid-level management, 10,000-49,999 employees, $10B+ revenue, consumer/retail
Source: 451 Research’s Voice of the Enterprise in-depth interview, July 2024.
Confidence in accuracy, and consequently trust, has declined with the rise of probabilistic models associated with generative AI. The proportion of respondents who distrust or only somewhat trust AI predictions remains elevated relative to 2023.
Previously, organizations had more trust in their homegrown capabilities compared with third-party models, but there is now greater alignment between the two.
Declining confidence that models will provide accurate outputs appears to be increasing interest in rules-based models — AI capabilities that use predefined rules and logic, which are inherently more structured than probabilistic models. Over a third of respondents (36%) believe that rules-based AI will be fully integrated in their organizations over the next 12 months, up from 20% who say the technology is already fully integrated.
“AI is just way too useful. Obviously, it has tons of flaws, and you have to constantly check. You cannot just take it at face value. But for finding similarities and causalities and little things, little nuances to large formats of data … it will find those things so much quicker than you could ever hope to.”
-IT/engineering manager/staff, 1,000-1,999 employees, healthcare
Source: 451 Research’s Voice of the Enterprise in-depth interview, November 2024.
When addressing challenges specific to generative AI, cost is frequently highlighted, along with data privacy, security risks, quality of generated content and integration challenges. As illustrated in Figure 5, many of these challenges can adversely affect project performance, diminishing the effectiveness of generative AI investments in achieving enterprise objectives. Although bias and fairness are perceived as less common concerns, organizations that have faced these issues often struggle to fully realize value from their generative AI investments.
Figure 5: Prevalence and severity of generative AI challenges
Q. Which of the following areas present challenges to your organization’s adoption of generative AI? Please select all that apply.
Q. How has generative AI affected your organization’s performance in the following objectives over the past 12 months? (Proportion of enterprise objectives strongly benefiting from generative AI among respondents who cite the given challenge)
Base: All respondents (n=999).
Source: 451 Research’s Voice of the Enterprise: AI & Machine Learning, Use Cases 2025.
Technological maturity is an issue, with research, capabilities and vendors evolving rapidly. This is a problem in isolation, but it is further exacerbated by skills shortages. Companies are not optimally positioned to maximize the value of generative AI technology due to a lack of skills in developing, managing and securing AI. A fifth of respondents report that cybersecurity skills shortages at their organization has caused project cancellation or failure. A further 32% say cybersecurity skills shortages have led to changes in project scope, delays or reduced value. More than three-quarters of organizations identified skills shortages in end-user expertise, data management and governance, cloud and infrastructure, software development and engineering, and machine learning and AI development roles. Without expertise, organizations will struggle to understand where the technology is best applied and how to effectively apply it.
“From the executive perspective, what they want … the solutions that are out there right now aren’t quite fully matured. For instance, they’ll look at AI to create this actual code. Well, 30% of the code that’s developed by AI is wrong. So, we’re not quite there at maturity level.”
-IT/engineering manager/staff, 500-999 employees, software, IT & computer services
Source: 451 Research’s Voice of the Enterprise in-depth interview, November 2024.
Outlook
Generative AI holds significant potential, with some organizations already recognizing its value as related technologies continue to advance. However, organizations’ initial investments are under the microscope, and many executives are less than impressed with the results. The generative AI pitch may need to pivot from the technology’s disruptive opportunity to a more concrete narrative about return on investment and demonstratable use cases. This changing pitch needs to address growing awareness of
generative AI’s cost implications, and it must consider how loss of enthusiasm may affect longer-term investment strategies. Regardless of these factors, results suggest that decision-makers will continue to press deployment aggressively over the next 12 months.
A more focused approach to generative AI will likely be guided by reassessed cost- benefit considerations, as well as by a technology transition toward AI capabilities targeted at specific tasks or workflows. Models tailored to specific high-impact challenges, integrated with applications and relevant data, may better justify organizations’ financial investments. However, as discussed in our companion report, “The next chapter in generative AI,” technology alone is insufficient. The new paradigm is not just about having generative AI, but about using it most effectively — a result that requires both technology and refined implementation strategies.
Further reading
Survey Data Hub - Voice of the Enterprise: AI & Machine Learning, Use Cases 2025
Generative AI experiences rapid adoption, but with mixed outcomes – Highlights from VotE: AI & Machine Learning
Voice of the Enterprise: Workforce Productivity & Collaboration, Employee Engagement 2024
The next chapter in generative AI: Agents, enhancement and the quest for value, February 2025
How to mastermind a successful AI project: Insight from IT professionals, November 2024
Moving from hype to ROI: Tracking success and value in AI initiatives, August 2024 Generative AI Market Monitor & Forecast, June 2024
Methodology
Voice of the Enterprise: AI & Machine Learning, Use Cases 2025 provides you with actionable data and insight and a broad, integrated view of enterprise AI/ML use cases, strategies and initiatives and their underlying business and technology drivers.
Combining 451 Research’s industry-leading analysis with insights from our extensive community of mid-level and senior IT and line-of-business professionals in North America and Europe, Voice of the Enterprise: AI & Machine Learning Use Cases 2025 provides evidence of real-world use cases across all industries, but with a particular emphasis on manufacturing, financial services, retail, healthcare, energy and telecommunications. This online survey of 1,006 respondents was conducted Oct.
21–Nov. 25, 2024. The margin of error for top-line statistics is +/- 3 pts at the 95% confidence level.
Note: Base sizes below n=30 should be interpreted anecdotally.
Note: Due to survey routing, qualification criteria, attrition and other factors, some questions were not answered by the full sample of respondents.
Demographics
For full survey demographics, click here to go to the Survey Data Hub.
About the author
Alex Johnston
Senior Research Analyst
Alex Johnston is a senior research analyst on the 451 Research Data, AI & Analytics team at S&P Global Market Intelligence. He focuses on emerging technologies and how they can be applied in business contexts. Alex’s primary coverage areas are artificial intelligence, distributed ledger technology and event stream processing.
Alex’s recent areas of concentration include monitoring the emerging generative AI market, tracking the evolution in blockchain use cases and investigating real- time architectures.
Before joining S&P Global Market Intelligence, Alex was a principal analyst at Procurement Leaders, leading much of the research on procurement and supply chain. His specialty focus was procurement and supply chain technology markets, advising chief procurement officers on digital roadmaps and building out a suite of diagnostic tools. He has a background in qualitative and quantitative research and has significant data science experience. Alex holds a history degree from the University of Warwick.