The rapid advancements in AI, particularly with Large Language Models (LLMs), have made the development of AI Online Shopping Agents (AIOSAs) a real possibility. While online shopping is a common task for many, a successful purchase often requires significant time and effort. For instance, finding the perfect desk for my 8-year-old daughter on Amazon took nearly an hour. I had to compare multiple products, understand their features, read customer reviews, and select the right size and color. Even after receiving the item, there might be additional tasks if I'm unsatisfied with the purchase.

In the foreseeable future, AI Online Shopping Agents can assist us in completing successful online shopping trips and save us time, energy, and money. For those who struggle with internet use, such as the elderly, the benefits would be even more pronounced.

However, developing an AI Online Shopping Agent is not a trivial task. It presents several challenges:

• Interacting with the Internet: AI must be able to interact with the internet as humans do. While recent advancements in AI agents, like Anthropic's computer use, have made strides, there's still room for improvement.
• Domain Knowledge: AI needs to possess domain-specific knowledge about shopping. It's not inherently knowledgeable about every aspect of shopping. Each step in the shopping process requires prompts to guide the AI, telling it which websites to visit, tools to use, and so on.
• Human Interaction: AI must effectively interact with humans, following human guidance and making decisions under human supervision.
• Safety and Security: With AI hallucinations still being an issue, ensuring the accuracy and security of AI-driven actions is crucial.

Given these challenges, we believe the development of AI Online Shopping Agents will be a gradual process. In the early stages, critical steps like placing orders may still require human-AI collaboration.

At BorderX Lab, we're actively researching and developing AI Online Shopping Agents. With a decade of experience in e-commerce and RPA services, we're confident in the potential of AI Online Shopping Agents and believe we're well-positioned to succeed. For example, our CloudStore AI product (https://www.nubestore.ai/) leverages RPA to automate order placement on major European and American fashion e-commerce sites with a success rate of 95%. This service can seamlessly integrate with AI to create AI Online Shopping Agents specifically for fashion e-commerce.