[Digital Today reporter Choi Jae-won (최재원)] A survey finds companies are introducing AI into their work, but fewer than 20 percent have achieved clear results strong enough to lead to organisation-wide innovation. The analysis says that while AI technology is advancing rapidly, the level of use in real corporate environments is falling short of expectations.

On Oct. 28, local time, IT outlet MIT Technology Review reported that a survey on the use of enterprise AI found all respondent companies were using AI, but only 16 percent said they had achieved innovative results. The survey was conducted by the outlet's Insight unit in cooperation with data management company Atlan, targeting 250 corporate executives in four regions including North America, Europe and Asia-Pacific.

40 percent of respondents said they achieved mid-level improvements in specific tasks or departments. 21 percent assessed their results as meaningful, but 23 percent said there was little or no effect relative to investment.

AI use also remained at an early stage. 38 percent of respondents said they were at the pilot stage of introducing AI, and 35 percent were using it to support employees' work, such as with copilots or chatbots.

Only 12 percent of companies said they had deployed AI agents in real work environments. Just 3 percent said AI performs most core tasks.

Experts pointed to a lack of understanding of work context as a reason companies are not fully using AI's potential.

Prukalpa Sankar (프루칼파 산카르), co-founder of Atlan, said there is a big gap between the capabilities of AI models and the level at which they are actually used in a useful way inside companies. She proposed a 'Context Layer' as a solution. It is a technical structure that connects internal company data and work systems to an AI platform and enables AI to understand organisation-specific rules and workflows.

For example, even with the same sales data, AI results can differ if sales and finance departments interpret it using different standards. The explanation says that if a company's work rules and data definitions are not properly conveyed, AI's judgments can become inaccurate.

Sankar stressed that more than AI adoption itself, the ability to systematically connect company data and use it to fit real work will determine future productivity and investment performance.