SK Telecom CEO Jaeheon Chung has introduced what he calls the “AX J-Curve,” arguing that companies seeking meaningful gains from artificial intelligence must first accept a period in which productivity can actually decline.
Speaking to about 300 undergraduate and graduate students at Seoul National University on September 1, Chung said AI transformation should not be viewed simply as adding new technology to existing workflows. Instead, organizations need to redesign the way work itself is performed — a process that can initially require more time and resources before delivering substantial productivity gains.
“Only when we move beyond familiar inefficiencies and confront extreme challenges at the edge can we reach the hyper-productivity innovation of the J-curve,” Chung said.
The lecture, titled “Choices and Opportunities in the AI Era,” opened the latest session of an AI joint program that SK Telecom and Seoul National University have operated for a decade.
Chung, who entered Seoul National University’s College of Law in 1987, returned to his alma mater as a lecturer nearly four decades later.
The ‘AX J-Curve’: productivity may fall before it rises
The centerpiece of Chung’s lecture was the “AX J-Curve,” a concept he developed from SK Telecom’s experience pursuing AI transformation over the past year.
The J-curve is traditionally used in economics to describe a pattern in which conditions initially deteriorate before improving sharply. Chung applied the concept to corporate AI transformation.
Companies often expect productivity to rise immediately after introducing AI. In practice, however, employee retraining, workflow redesign, data preparation and organizational changes can initially make operations less efficient.
Organizations that endure this transition and fundamentally redesign their workflows around AI can eventually reach a tipping point where productivity begins to rise rapidly, according to Chung.
“If you turn back at the valley of discomfort, you remain forever on the left side of the curve,” he said. “Organizations that endure it eventually turn upward in a J-shaped trajectory.”
Chung identified three principles for crossing that valley: looking at familiar work from a new perspective, confronting extreme challenges and accepting temporary discomfort.
From 50 hours to 500 — and ultimately one hour
One example from inside SK Telecom illustrates the productivity paradox.
According to Chung, an SKT developer previously spent about 50 hours handling a system modification requested by customers. When the company attempted to automate the process with AI, the initial effort required more than 500 hours.
The additional time was needed to build an AI agent, modify it and conduct hundreds of rounds of validation.
After completing that process, however, the same type of task can now be performed in about one hour.
The example highlights a central argument behind Chung’s J-Curve: measuring AI solely by its immediate efficiency gains can obscure the larger productivity improvements that become possible after workflows are rebuilt around the technology.
SK Telecom is now applying a similar principle more broadly by asking a fundamental question across its operations: “If AI were the starting assumption, how would we redesign this work?”
Sovereign AI emerges as a strategic issue
Chung also placed corporate AI transformation within the broader global competition over AI capabilities.
He argued that AI is increasingly becoming a strategic asset not only for companies but also for countries. Depending excessively on technology developed elsewhere can leave organizations with limited control over pricing, access and operating conditions, he said, underscoring the importance of sovereign AI capabilities.
SK Telecom has been pursuing its own AI model development as part of that strategy. Chung cited the development of A.X K2, a 688-billion-parameter large-scale AI model built under constrained GPU resources, as an example of setting goals that cannot easily be achieved through conventional approaches.
For Chung, such constraints can themselves become catalysts for innovation by forcing organizations to abandon familiar methods.
AI is changing the ‘density of time’
Chung described AI as a technology capable of changing what he called the “density of time.”
While everyone technically has the same 24 hours in a day, he argued that AI could effectively give some individuals the productive capacity of 240 or even 2,400 hours by dramatically expanding what can be accomplished within the same period.
The speed of AI adoption is also challenging assumptions formed during previous technological transitions, he said, as generative AI services have spread much faster than technologies such as automobiles and mobile phones.
That acceleration means companies can no longer treat AI as another incremental IT upgrade. Instead, Chung argued, it requires organizations to reconsider established assumptions about which tasks humans should perform and how work should be structured.
Human judgment becomes more important, not less
Despite emphasizing AI-driven productivity, Chung rejected the idea that AI will eventually make all important decisions.
He said the ability to set direction, exercise judgment and understand other people will become increasingly valuable as AI takes over more technical tasks.
Individuals should develop deep expertise in their own domains and use AI to amplify their productivity, he said. Those capable of crossing traditional boundaries between technology, industries and people could then create entirely new opportunities.
Chung also referred to SK Group Chairman Chey Tae-won’s four qualities for talent in the AI era: the ability to think deeply about fundamental issues, adapt and learn from failure, empathize with others, and turn accumulated physical and practical experience into value.
“None of these four are coding skills,” Chung said. “Technology is something AI will increasingly become better at, which is why the capabilities unique to humans become the conditions for talent.”
He concluded by contrasting today’s AI debate with the questions being asked a decade ago.
“Ten years ago, we asked whether AI could surpass humans,” Chung said. “Today, we are asking what we will create with AI.”
Ultimately, he added, judging what has value and deciding which direction to pursue will remain a human responsibility.
