Can you believe it? Half of AI models never make it to production

​The recently released Gartner 2022 AI in Organizations survey found that 80% of executives believe automation can be applied to any business decision. As automation becomes increasingly important to business operations, the survey explores how businesses are embracing the use of artificial intelligence as part of their long-term automation strategy. The report estimates that the commercial value of artificial intelligence will reach $5.1 billion by 2025.

Gartner associate analyst Erick Brethenoux said: “The survey shows that enterprises are moving away from adopting artificial intelligence in a purely tactical , moving towards more strategic applications of AI.” “For example, one-third of organizations are applying AI across multiple business units to create stronger competitive differentiation by supporting decision-making across business processes.”

The survey found that an average of 54% of artificial intelligence projects are making the leap from experimentation to production, higher than the 53% in the 2019 AI in Organizations survey report. Although this number has not increased significantly, the proportion of deployed AI projects has increased from 35% in 2019 to 47% in the latest survey. Gartner attributes this to a more flexible approach to automation brought about by the pandemic, which allows for more practical implementation of AI within organizations.

But scaling artificial intelligence remains a major challenge. Enterprises are still struggling to connect the algorithms they are building to business value propositions, and it is difficult for IT and business leaders to justify the investment required to operate the model.

One issue with scaling AI is the complex governance required when deploying large numbers of AI models, with 40% of organizations surveyed reporting deploying thousands, sometimes even hundreds of thousands, of AI models. More models create governance challenges that can make it difficult to demonstrate value and ROI, a difficulty seen as the biggest barrier to AI adoption.

The survey also found that cumbersome barriers to entry, including finding the right talent, have been reduced. Despite reports of talent shortages in other technology areas, Gartner's survey found that the lack of AI talent is not a significant barrier to AI adoption for organizations surveyed, with 72% of executives saying they have or can find talent for their projects. Required AI professionals.

The most successful organizations use a combination of internal selection and external recruitment of AI talent. This ensures that the team is constantly updating itself by learning new AI skills and technologies, and considering new ideas from outside the organization.

Security and privacy concerns were also not seen as the biggest barrier to AI adoption, with only 3% of respondents citing this, although 41% disclosed previous AI privacy breaches or security incidents . While not the most significant obstacle to AI, it is still a concern. Half of the respondents said they were worried about competitors, partners or various third parties gaining access to sensitive information, and 49% were also concerned about malicious external attacks. Interestingly, 60% of respondents who faced an AI security or privacy incident said their data was compromised by an internal party rather than an external source.

“Given that most AI intrusions are caused by insiders, organizations’ concerns about AI security are often misplaced,” Brethenoux said. “While attack detection and prevention are important, AI security efforts There should be an equal focus on minimizing human risks."

Gartner conducted the survey online from October to December 2021. The 699 respondents from the United States, Germany, and the United Kingdom are companies that have deployed or plan to deploy AI within three years. Gartner analysts also discussed the findings at the Gartner Data & Analytics Summit in Orlando on August 24. ​

The above is the detailed content of Can you believe it? Half of AI models never make it to production. For more information, please follow other related articles on the PHP Chinese website!