GenAI's experimental phase may be over as companies reduce experimentation and shift to realizing business value, focusing on fewer, more targeted use cases.

According to a recent survey by NTT DATA, nearly nine in 10 senior decision makers say they are tired of the GenAI pilot and are shifting investments to projects that will improve business performance.

Andrew Wells, NTT DATA's chief data and AI officer for North America, said businesses will continue to experiment with new GenAI pilots, but a more targeted approach that focuses on their business-specific use cases is increasingly becoming central to IT leaders' AI strategies.

Wells said that in some cases, pilot failure rates as high as 50 percent or higher are forcing companies to rethink the number of pilots they launch. An IDC survey in April found that, on average, companies have launched 37 proof-of-concept AI projects, but only a handful have made it to the production stage.

"Either you don't have the right data to do it, the technology isn't mature enough, or the model doesn't exist yet." Wells explains why the early trials failed so often.

In other cases, he says, pilot projects are not commercially viable. "You build proof-of-concept projects, but the effectiveness of the solution doesn't necessarily match the initial assumptions." "He added.

In addition, more than a third of IT professionals say that the goal of the AI projects they are involved in is not practical value, but to show investors and stakeholders that their organization is working on AI.

Wells said a large number of GenAI pilots that have not progressed are consuming resources.

"When we go into most companies, they have a significant backlog of GenAI use cases, hundreds to be specific," he said, "and they are now more purposefully deciding what projects they want to invest time, effort and money in, rather than 'let's try it first and see what this technology can do.'"

undercost

Courtney Schuyler, CEO and co-founder of SkyPhi Studios, a digital transformation consulting firm, said companies often launch AI pilots without considering hidden costs.

Launching multiple pilots in a short period of time is not only costly, but often results in reduced employee productivity as they struggle to learn how to use the new technology.

"What I often see is companies rushing into this technology boom without really slowing down, thinking strategically about how to leverage this technology within their enterprise, what strategic path to take, and just looking at the overall change fatigue that organizations may be experiencing based on other technology implementations," she said.

Schuyler says moving the project to the implementation stage is only part of the battle. Getting employees to adopt the new GenAI tools was a huge step.

"It's important to be really aware of your investment and the costs associated with it, from the cost of the technology to the cost of helping people adopt it," she adds. "From my perspective, it's more important to take the time to look at and plan a budget for this than to rush into it and potentially lose millions of dollars, because if it doesn't work out as expected, then the losses will be huge."

Spending is still rising

Despite mixed results over the past year, many companies are planning to increase spending on GenAI in 2025 and beyond. In the NTT DATA study, 39 percent of respondents said they now have a significant investment in GenAI, a percentage that will rise to 61 percent in the next two years.

NTT's survey aligns with the findings of a new survey commissioned by IBM, which found that 62% of companies plan to increase their AI budgets by 2025. Still, despite pilot fatigue, more than a quarter of respondents said their companies plan to launch more than 20 AI pilots by 2025.

Dev Nag, CEO of pro-autonomy firm QueryPal, said that while experimentation will continue, many businesses are likely to focus on projects that give them a competitive edge, rather than general-purpose HR, digital assistant, or chatbot projects.

Nag said that as early AI projects, many IT companies tried to create their own chatbots and HR AI, but now some companies are outsourcing these functions to AI vendors.

"We're seeing this on a micro level, where people are trying to build [chatbots] themselves with teams that aren't dedicated to support and don't have any background in AI," he said, "and that's creating a huge overinvestment." We've made companies almost become venture capitalists, funding IT projects as if they were start-ups."

But Nag says it doesn't make sense for every business to build its own HR AI.

"Do you really want 10,000 companies trying to build customer support agents, HR agents and finance agents?" 'he said.

Instead, he added, many businesses appear to be turning to a smaller number of GenAI pilots focused on their unique needs, rather than general-purpose chatbots.

"For most companies, if it's not particularly relevant to your bottom line, then it's a distraction and a failure," he said. "You lose people because they get thrown into something like this and it turns out, 'We thought this was going to be a successful center.'"

Specific demand

Aaron Schroeder, director of analytics and insights at contact center IT provider TTEC Digital, sees some of the same trends. Most of GenAI's major advances in public outreach have come from a generic model focused on a single use case, rather than complex commercial uses, he said.

"These models and functions are based on broad knowledge from the Internet, rather than a specific domain and context," Schroeder added, "This has allowed many leaders to see how emerging AI solutions can help them in their daily lives, but there is still a gap between that and seeing AI solutions deliver meaningful productivity in hyper-specific, industry-oriented use cases that require an understanding of how the company operates."

That gap, he said, is prompting companies to shift pilots from general-purpose programs to ones that are more aligned with value-driven areas.

"The most successful approach we've experienced is to design the governance of AI pilots and solutions ahead of time at a higher level - whether to accelerate productivity, save costs, increase revenue, or enhance the customer experience," Schroeder said. "By identifying key principles ahead of time, it's easier to maintain consistency across multiple simultaneous projects."