불가능의 자동화 : AI 커머스 스타트업 인핸스의 Vertical AI Agent 구현 사례

Summary

Dr. Kim Do-gyun, CAI CAIO of Ais, presents the development of AI-powered e-commerce agents aimed at automating complex tasks in product management, pricing, and market analysis. His company, leveraging AWS services, has created autonomous agents that enhance operational efficiency and scalability by automating repetitive functions and data collection. With over 10 billion won in sales, Ais has pioneered AI in commerce, using agents to handle long-tail e-commerce platforms. The company’s future goal is to automate agent creation and management, relying on a comprehensive knowledge database to drive innovation and performance.

Takeaways

• 😀 Dr. Kim Do-gyun, CAI CAIO of Ais, presents on the implementation of AI agents in e-commerce automation, focusing on ethical AI and scalability.
• 😀 Ais believes that by 2030, AI will take on roles on both the seller and buyer sides in e-commerce, reducing human intervention in transactions.
• 😀 Ais has already generated over 10 billion won in sales through AI agents, scaling up and selling these agents as a service.
• 😀 E-commerce faces challenges such as data utilization, human involvement in operations, intense price competition, and the lack of comprehensive solutions in Asia.
• 😀 AI agents in e-commerce can help automate functions like market research, price adjustments, and ad analysis, making real-time responses possible during off-hours.
• 😀 Ais transitioned from unit function development to AI agent development to address limitations of scalability, human errors, and rigidity in their systems.
• 😀 The company has automated functions like display analysis and price adjustments, allowing agents to collect and analyze data, making reports for admins automatically.
• 😀 Large-scale utilization of Language Models (LM) like LLMs has raised concerns about frequent updates, API security, and the need for a robust infrastructure to support AI agents.
• 😀 AWS services played a pivotal role in building scalable, flexible AI agents, including automated web agents that interact with e-commerce sites, adjusting dynamically to peak traffic periods.
• 😀 The goal of Ais is to automate the creation of agents, develop APIs, and build a knowledge database on commerce, eventually automating agent management and creation workflows.

Q & A

What is the main goal of Dr. Kim's company?

The main goal of Dr. Kim's company is to automate e-commerce processes using AI agents, ultimately aiming to become an AI-powered agent company that can create and manage workflows autonomously through a knowledge database and APIs.

What vision does Dr. Kim have for the future of e-commerce?

Dr. Kim envisions that by 2030, most transactions in commerce will be conducted through AI, with little human intervention, and AI agents will play a key role in driving both the buyer and seller sides of the market.

How does the company use AI agents in e-commerce?

The company uses AI agents to automate various e-commerce tasks, including analyzing competitor strategies, optimizing display advertising, and managing inventory. This automation allows agents to operate 24/7, handling tasks that would otherwise require human intervention.

Why did Dr. Kim’s company decide to shift from task automation to AI agents?

The shift was necessary because task automation alone lacked scalability. Initially, the company automated single tasks, but when platforms changed or new requirements arose, manual intervention was needed. AI agents, capable of recognizing goals and adapting to dynamic environments, provided a more scalable solution.

What is the function of the display agent developed by the company?

The display agent automatically analyzes search results, compares competitors’ advertising strategies, and generates reports. It is designed to monitor over 300 e-commerce sites across 53 countries, helping clients stay competitive without manual oversight.

How does AWS help Dr. Kim's company scale their solutions?

AWS provides the infrastructure necessary for scaling the company’s AI solutions. For example, EC2 helps manage dynamic instances based on traffic, and DynamoDB offers a flexible database solution. The use of AWS’s cloud services enables the company to deploy models quickly and scale them as needed.

What role does Bedrock play in the company’s AI strategy?

AWS Bedrock is crucial for improving the reasoning capabilities of the company’s AI models. By using Bedrock, the company can update and improve its planning models more efficiently, allowing AI agents to respond faster to changes and new data in real-time.

What is the Large Axis Model (LAAM) framework, and how does it contribute to the company’s AI solution?

The Large Axis Model (LAAM) framework is used to create autonomous web agents that can gather data from hundreds of platforms in real-time. This model enables the company to provide valuable insights and prevent missed opportunities in the long tail of e-commerce, improving competitiveness.

How does the company plan to further automate their processes in the future?

The company plans to automate the entire process of creating and managing AI agents by building a knowledge database and ontologizing commerce knowledge. They aim to automatically generate workflows and create agents without human input, making the entire process autonomous.

What are some of the challenges Dr. Kim’s company faces when working with AI models?

One of the main challenges is keeping the models up to date and managing the complexity of frequent updates and version changes. This is particularly important as the company operates across multiple platforms with diverse requirements and data sources.