Share with your CEO
Wrtn Technologies, a Seoul-based AI platform that aggregates ChatGPT, Gemini, and Claude into a single interface, is betting that AI transformation fails when it starts with the easy stuff. Park Min-jun, the company’s head of AI transformation, told Korean industry leaders at the FKI CEO Jeju Summer Forum that the correct entry point is revenue-driving work, not administrative repetition. He also claimed Wrtn has already replaced “a significant portion” of executive decision-making functions with AI since February 2026, with working-level staff next.
What this means for your business
The companies getting the least from AI right now share a common pattern: they automated expense reports and meeting summaries first, declared victory, and stopped. Park’s argument inverts that sequence entirely. If your AI program is two years in and still living in back-office workflows, the question worth asking isn’t whether AI is working, it’s whether you ever pointed it at anything that actually drives revenue. That’s the test this framing puts to every CEO currently reporting “broad AI adoption.”
The three prerequisites Park names, internal coding capability, machine-readable documents, and cultural change, sound procedural until you notice the ordering. Most organizations invert the third and treat culture as the final polish rather than the prerequisite that determines whether the first two ever get used. The hackathon model Wrtn describes, non-developers first, bottom-up enthusiasm rather than top-down mandates, is a real counter to the “AI as extra homework” resistance pattern. Mandated adoption programs consistently produce compliance theater; they don’t produce the “wow, this actually works” moment that generates genuine behavior change.
The claim that Wrtn has replaced executive functions with AI since February is either the most interesting data point in this story or the most slippery one. “Replace” almost certainly means augment specific decision workflows rather than eliminate roles, and Park’s own framing, that his personal decision speed and quality improved, suggests the latter. But the direction of travel matters regardless. The organizations that will feel this most acutely aren’t the ones still debating whether to pilot AI; they’re the ones that have piloted it and let it idle in a department that doesn’t touch a revenue line. That’s the budget conversation a CEO should already be having with their CIO.
Concept deep-dive: Action agents
An action agent is an AI system that doesn’t just answer a question but executes a task in an external system, think of it as the difference between a navigator reading you the map and one who actually turns the wheel. Canceling a subscription, rerouting a delivery, and issuing a coupon are Park’s examples. The business significance is that action agents introduce real operational liability: a wrong answer is correctable, but a wrong action already happened. That asymmetry is what makes governance the non-negotiable design layer, not an afterthought.
Based on reporting from AI transformation must start with revenue-driving work, not routine tasks, Wrtn Technologies executive says, originally published 2026-07-16 22:14:00.
