As agent-type AI evolves beyond simple question-and-answer interactions to actual task delegation and deliverable generation stages, workflow restructuring in knowledge-intensive sectors such as research and media is becoming tangible. OpenAI publicly demonstrated for the first time the domestic application potential of ChatGPT Work, an agent service based on the Codex coding agent.

 

On the 20th, OpenAI held a 'ChatGPT Work Media Demo' at its Seoul office, demonstrating key features and revealing actual work application cases from domestic experts.

 

Codex surpassed 3 million weekly active users (WAU) globally in April of this year, and following ChatGPT Work's launch in July, the combined WAU of both services is reported to have exceeded 10 million.

 

Shim Dae-yeol, OpenAI Korea engineer, demonstrated the integration structure of ChatGPT Work on the day.

The method involved retrieving contextual information from external work tools such as Slack and Google Drive through plugins to generate deliverables like presentations, and regularly updating project information through scheduled tasks.

 

Ko Gil-gon, director of the Information Strategy Office and professor at the Graduate School of Public Administration at Seoul National University, introduced a case of establishing a research workflow that collects and structures statistics, laws, policy documents, and academic materials by country using Codex and connects to draft reports.

 

Lee Kyung-hee, head of the Content 1 Division at JoongAng Ilbo, announced a case implementing a reporting dashboard that standardizes and verifies public official asset disclosure data to enable personnel and institution search, year-on-year comparisons, and data change verification.

 

Um Seong-won, OpenAI Korea Communications Director, said, "Usage methods are expanding toward delegating actual work to AI and having people focus on more important decision-making and problem-solving."