Ask HN: How can AI help as a shopping assistant?

2 pointsailef2y ago2 comments

Hi HN,

I'm developing an AI-based shopping assistant and I was looking at some feedback on the idea.

Assumption: When somebody needs to buy a high-value item there's usually some research involved in finding the right fit and comparing different products. My idea would be to provide value by simplifying this process.

This is a short video demo of the current state of the tool: https://www.youtube.com/watch?v=C5KVqcfQpB8 (it's a WIP and there are still some rough edges).

The way the tool works is it asks some questions in order to refine the search, and then evaluates the products based on your answer. After ranking, the products are shown and the requirements are marked as satisfied or not. There's also a Chat with AI feature where you can ask questions about the products and its reviews, if you need extra information.

There are other features I have in mind, like product comparison, but I'm slowing down a bit and looking for some feedback before developing something that nobody wants.

At the moment it's limited to Amazon.com, but the plan would be to extend support for other online retailers (which ones? I'm targeting US market but I'm not sure which would be worth to add, if any).

Do you think it is viable? Is there any value in this or you prefer to just search yourself on Amazon.com (maybe this holds for the HN demographic, but what about "normal" people)? What kind of feature would you like to see?

These are the kind of questions I'm asking myself, so if somebody wants to chime in and give their two cents I'd appreciate it very much.

Thanks!

2 comments

haebom2y ago

At the end of the day, consumers are going to look for the cheapest and fastest product they can get. E-commerce in China, e-commerce in Korea, Amazon Prime did it, eBay didn't do it, and now consumers want to get what they need cheaply and quickly. Personally, I'd rather use good old-fashioned methods than AI for this. It's too expensive to spend LLMs on search, product reasoning, and recommendations right now (and it's slow!).

ailefOP2y ago

These are all fair points I agree with. Let's just say I might've started writing code without thinking too deeply about the problem...

About the last point (slowness), this is definitely a problem right now. In my case, also retrieving the information from Amazon takes a sizeable amount of time; for the LLM part I think it can be optimized with finetuned models and performance will generally get better in the future. So maybe overall slowness is the easiest problem to solve.

j / k navigate · click thread line to collapse