Early adopters report major benefits, but 28 percent of AI projects still fail

More than a quarter of all AI initiatives have already moved into production, with more than a third in advanced development stages, according to a new study by International Data Corporation (IDC).

More than half of large companies reported they were using AI to deliver a better customer experience.

The majority of 2,100 IT and line-of-business decision makers surveyed said that the greatest impact of AI on their organizations was observed in helping employees get better at their jobs.

Whether an improved customer or employee experience, the study found a direct correlation between AI adoption and superior business outcomes.

"Early adopters report an improvement of almost 25 percent in customer experience, accelerated rates of innovation, higher competitiveness, higher margins, and better employee experience with the roll-out of AI solutions,” said Ritu Jyoti, program vice president for AI strategies at IDC.

“Organizations worldwide are adopting AI in their business transformation journey, not just because they can, but because they must, to be agile, resilient, innovative, and able to scale.”

No longer nice-to-have

The leading use cases for AI include IT automation, intelligent task/process automation, automated threat analysis and investigation in cyber security, supply and logistics management, automated customer service agents, and automated HR operations, according to IDC.

Automated customer services agents and automated HR are a priority for companies with 5,000 or more employees, while IT automation is the priority for companies with 1,000 or fewer employees.

Businesses also face several challenges in AI deployments. A significant development roadblock is the lack of adequate volumes and quality of training data.

The leading data integration challenges were found to be in security, governance, performance, and latency or transfer rate, and the top data management issues are the price of solutions, performance, and scale.

However, the overall leading challenge for enterprises implementing AI remains cost, with fragmented pricing across different services and pay-as-you-go pricing presenting barriers as enterprises scale their efforts.

Enterprises also reported they spend a third of their AI sytems’ lifecycle on data integration and data preparation, rather than actual data science efforts.

“Businesses will need to embrace machine learning operations, the compound of machine learning, development, and operations to realize AI/ML at scale,” the report stated.

Nearly a third (28 percent) of AI and ML initiatives were reported to have failed, primarily due to lack of staff with the necessary experience, lack of production-ready data, and lack of integrated development environment, according to IDC.