[AI for All, Divergent AX]②Experts Discuss the Causes of AX Failure

Why Is AX Implementation So Challenging for Small and Mid-Sized Businesses?

Editor's NoteArtificial intelligence (AI) is transforming corporate productivity and the way people work, but the speed of change differs between large corporations and small and medium-sized enterprises (SMEs). While companies with dedicated AI teams and infrastructure are accelerating company-wide AI transformation (AX), those without such resources—both businesses and organizations—are falling into AX blind spots, left out of the promise of "AI for all." The Asia Business Daily examines the AX divides that arise beneath the surface of innovation, and explores the conditions and solutions necessary to raise productivity through truly inclusive AI that leaves no one behind.

"While manufacturing conglomerates can invest around 1 to 2 billion won in the same financial automation projects, SMEs find it difficult to even invest 30 million won."

Companies that support their clients' AX initiatives frequently encounter these "AX blind spots" in the field. In interviews with The Asia Business Daily on August 12, experts warned that the AX gap between large companies and SMEs is widening, particularly in terms of each group's "readiness" and "execution capabilities" to connect AI with work and outcomes. They cautioned that, without government support, AX polarization is likely to intensify.

Park Minjun, CEO of Ruitton AX, explained that in practice, once large corporations commit to AX, they undertake projects of over 1 billion won per department and maintain ongoing investments. In contrast, SMEs often view even the upfront costs for a single department as a significant burden. The budget difference immediately translates into disparities in AX levels. Plans for infrastructure or capabilities often have to be scaled back, and customized services tailored to each work environment are frequently abandoned. Park noted, "Among SMEs, it is common to abandon their intended AX level due to cost concerns. If available investment is limited, even custom projects become challenging."

Gong Seongbae, Chief AI Officer (CAIO) at Megazone Cloud, explained that the AX gap observed in the field goes beyond a simple difference in "AI adoption rates." Although generative AI has quickly become widespread, ostensibly allowing everyone to use AI, clear differences remain in how organizations apply AI to their actual work and spread it throughout the organization.

Gong observed, "While large enterprises experiment with various AI projects and consider company-wide integration and unified governance, the majority of companies remain at the level where individual employees use AI independently or are just beginning to identify which work areas to apply it to." He added, "Unlike large businesses, which have AX execution teams and budgets in place, it is common for just one or two staff members to juggle both their regular work and AX-related tasks at the same time."

The situation is similar not only for SMEs, but also for small business owners who have not experienced AI-driven economic benefits, public institutions with strict regulatory and security requirements, and educational organizations where data, mostly in Korean documents and paper format, is poorly standardized. Gong added, "Many organizations struggle to secure specialists, and regulations make it hard for them to use open AI services. The gap between early and late adopters could widen over time."

One thing that organizations failing at AX have in common is a lack of sufficient usable data. Their data is scattered across various systems and documents, with differing standards and quality, and access rights are undefined, making it difficult for AI to leverage the data. When work relies heavily on employee experience and informal collaboration, it is also difficult to build expert domains.

Han Seonho, CAIO of Bespin Global, said, "To get agentic AI to function properly, organizations must first modernize their legacy systems and structure their data. Yet many businesses lack adequate budgets and personnel for this preparation phase."

Another issue is the absence of a clear internal roadmap for achieving tangible results with AI. Han added, "Many companies stall at the proof-of-concept (POC) phase because they lack the in-house capacity to determine which work processes and methods, and in what order, to apply AI to."

Experts unanimously agree that bridging the AX divide requires more than simply providing solutions or one-time assistance; support must extend to data organization, consulting, training, and post-implementation operations.

Park emphasized a "triangle model" of education, consulting, and implementation to achieve customized AX. He explained, "AX is not just about building and selling a chatbot—it involves changing the overall corporate culture and reaching internal consensus on the need for AI transformation. The first areas to transition to AI should be those that generate real revenue and reduce costs."

There is also a need to create standardized AX criteria by industry and sector. It is essential to establish a foundation for jointly utilizing data, computing resources, and specialized talent that individual organizations find difficult to secure on their own. Park said, "If the needs of organizations with limited investment capability can be gathered and standardized, and then matched with solution providers, the pace of AX can be accelerated."

This content was produced with the assistance of AI translation services.