[Edaily Reporter Kim Hyun-ah ] Korea Deep Learning is conducting a pilot project in collaboration with SKTelecom(017670)to automate processes ranging from product information analysis to the production and review of advertising videos using artificial intelligence (AI). This involves applying visual language model (VLM) technology—previously used for document processing—to commerce content production tasks.

On the 22nd, Korea Deep Learning announced that it is conducting a pilot project to automate product information analysis and advertising production using its enterprise Agent AI, “DEEP Agent.” This collaboration is being carried out as part of the “SKTelecom Startup Open Innovation Challenge 2026” project organized by the Busan Creative Economy Innovation Center.

Analyzing Product Information to Create Advertising Videos

Korea Deep Learning has configured a system that uses VLM technology to analyze product names, prices, images, and other details from online store product pages, and to classify and select the information needed for ad production. The extracted information is used to craft ad copy and compose scenes.

The produced videos undergo automatic review based on criteria such as consistency with the original product information, readability of the copy, and missing images. If a video fails to meet the review standards, it is either revised or regenerated.

During the interim validation phase, 12-second short-form advertising videos were produced for three categories—women’s clothing, food and beverages, and toys—based on 11st product pages. The team tested the automated process, which spans from product information collection to image selection, scene composition, output review, and video generation.

Beyond Document Processing to Content Creation

This pilot project is significant in that it applies AI technology—which extracts and analyzes information from documents—to the production of e-commerce content. The goal is to verify whether AI can go beyond simply analyzing information to also handle follow-up tasks and review the final output.

Founded in 2019, Korea Deep Learning has been developing enterprise document automation solutions based on visual language models and optical character recognition (OCR) technology. Through this pilot, the company plans to verify whether AI agents can be utilized in tasks other than document processing, such as content creation.

Kim Ji-hyun, CEO of Korea Deep Learning, stated, “We are exploring the potential for AI to understand the context of product information and images and to link this to actual production and verification tasks,” adding, “We plan to verify its applicability across various industries and work environments.”