Infofla to launch AI Agent for Comprehensive Online/Offline Task Automation.

- Two AI-based task automation solutions, VLAgent and FLAR,  to overcome 
  the limitations of traditional automation tools.
- VLAgent can interpret both natural language and on-screen visual elements
- FLAR extends automation into the physical world by enabling  intelligent 
  control of robots

Amazon Web Services (AWS), a global leader in cloud computing, held its annual AWS Summit Seoul 2025 at COEX from May 14th to 15th. As the largest cloud event in Korea, the summit featured a wide range of AWS’s cloud and AI-related visions.

In addition to AWS-led sessions, several AWS partners showcased their technologies and products. Among them, one of the impressive keynotes was delivered by Infofla, a company specializing in AI-based task automation solutions.

Choi In-mook, CEO of Infofla / Source=IT Donga

During the session, Infofla CEO Choi In-Mook emphasized the need to overcome the limitations of existing automation solutions through AI. While traditional Robotic Process Automation (RPA) tools have automated repetitive tasks, they typically follow rigid patterns and struggle with unexpected changes.

For example, when browsing a website, if a popup window appears or a screen element shifts position, conventional automation tools often fail and stop functioning. Although some Large Language Model (LLM)-based agents have attempted to address this, they are generally limited to automating text-based web pages. To overcome this limitation, Infofla introduced two key solutions:

• VLAgent – Vision-based AI Agent
• FLAR – Vision-based Robotics Foundation Model

Vision-based AI Agent That Understands Screen: VLAgent

VLAgent can interpret both natural language and on-screen visual elements such as images. For instance, if a user inputs “Show me how to get from Konkuk Univ. Station to Gangnam Station,” VLAgent can automatically move the mouse and keyboard to launch a Chrome browser, open a mapping service like Naver Maps, enter the departure and destination, and return the search results. Unlike typical RPA tools, which may stop functioning due to a single popup, VLAgent can automatically close such popups and continue its task seamlessly.

VLAgent system architecture for task automation. / Source=IT Donga

CEO Choi emphasized that ease of use is one of the major strengths of VLAgent. Developed using open-source models, the solution can automate tasks both in cloud environments and in closed, offline networks that are disconnected from the internet. Even normal non-technical users can train the system to learn the new processes simply by clicking a “Record” button, performing their specific workflow, and then clicking “Save”.

Automation the Physical World: FLAR

Infofla also revealed plans to launch “FLAR,” which is designed to extend automation beyond digital interfaces into the physical world. By enabling intelligent control of robots, FLAR opens the door to far more comprehensive automation across industries. In a retail environment, for example, the system can analyze next-day product schedules after business hours and autonomously instruct robots to stage and prepare inventory in advance which which helps streamline operations and reduce manual labor.

Infofla develops task automation covering both online and offline environments. / Source=IT Donga

Infofla is developing FLAR, in collaboration with the Electronics and Telecommunications Research Institute (ETRI) in Daejeon and Stony Brook University in New York. The company aims to complete development by next year.

Proven Results and Global Expansion

Infofla’s solutions, including VLAgent, have already been successfully deployed in several organizations including Korea’s Ministry of the Interior and Safety (MOIS), a local government agency in Japan, the e-commerce platform DaisoMall, and so on.

In particular, government agencies like the Ministry of the Interior and Safety (MOIS) need to inspect and review thousands of public websites to comply with the Korea Design System (KRDS), a national initiative recently introduced to standardize public website design. According to INFOFLA CEO Choi In-Mook, Infofla’s automation solutions are delivering remarkable results in handling that such kind of large-scale repetitive workload.

Additionally, CEO Choi also announced that Infofla is in the process of establishing a U.S. subsidiary to support international expansion. The company is working closely with AWS to broaden its global services and partnerships.


By Young-woo Kim (pengo@itdonga.com)