Botify

Early e-commerce shops suffer from low level of user activity to collect user behavioral data to train their search/recommendation models, gain data driven insights of customers. By deploying chatgpt agents, prompt-engineered with a profile that describes a imaginary user, early start-ups can collect user logs from multiple fake agents exploring their website. In our demo, the "agents" is where users can create a fake user giving rough information of an imaginary user with age, location, interests and description. Upon clicking the new task, it dispatches the fake user to interact with a shopping mall (in our demo we used amazon), and the agent with search, click recommended items, ponder around reviews just like the ordinary user would do when coming to a conclusion to buy a product. The "logs" page is a list of behavior logs that is crucial in training search/recommendation models, as it can become a train dataset to find relationship of query-to-product (search) and product-to-product (recommendation). Additionally, data scientists can analyze the logs to gain insights from user behaviors. Our project Botify is dedicated to solve the chicken-and-egg problem in e-commerce, where they first need user data to sophisticate their algorithm for better user experience. While the demo is yet preliminary proof of concept, we hope to see much more development when properly used with advanced prompt engineering techniques