A-Insight

Inspiration

It came when we talked to brand owners on how they navigate through and operate their e-commerce store. They rely on their account manager or analyst to know current industry trends, whether their store is doing well or not, and etc. They are presented with a dashboards that has a lot of metrics and little understanding of any insights they could gain from it.

They need to go the extra mile to extract data from dashboards, understand them and create insight from those data. Once the insights are clear, then they develop an action plan. However, the process took so long and was very inefficient. We see these gaps and try to fill in using AI.

What it does

A-Insight provides you actionable insights from your online store dashboard. It works like chatgpt, where you ask several prompts and it provides u insights about your store. It can provide you with actionable insights in seconds on how to improve your store sales, identify which product has low visitors and needed change of product catalogues, how to increase traffic and conversion rate, and so on. With an easy to navigate interface and providing actionable insights in seconds could significantly improve store sales and increase productivity.

How we built it

We built A-Insight by integrating AI into our process. By incorporating e-commerce data into our AI, e-commerce sellers can now talk to the AI about their shop performance and gain actionable insights. Additionally, we will focus on our AI's specialties by providing it with the latest trends and knowledge about e-commerce and digital business.

Challenges we ran into

Gaining access to e-commerce data might invade user privacy. However, less data means the model might not learn as much. It might reach a dilemma and we need to find another approach on how we can get trusted, consented and reliable data to train our ML Model.

Accomplishments that we're proud of

Actually, we haven’t developed this, but we talked to many potential users which are brand owners that interact with e-commerce dashboard everyday. When we pitch this idea to gain new insights, they immediately want to consider themselves early adopters and willing to try as soon as the beta is launched.

With this solution, they felt that it should increase their productivity and gaining insights that they might not have covered at the first place

What we learned

Generative AI has many models and capabilities, which also affect its output and pricing. From this competition, we learned that selecting the model that best suits our needs is necessary. For our purposes, we chose to use GPT-4o because of its capabilities to process multimodal input and its fast response time.

GPT-4o also has the capability to process and visualize data, which matches our need to integrate e-commerce data, usually in the form of spreadsheets, and create visualizations that can be turned into actionable insights for the seller.

What's next for A-Insight

Gaining access to the data might be an impossible task. However, if we managed to gain real-time data and it was reliable and only used for insight-driven, it might increase the accuracy and reliability of the actionable insights that the AI might provide.