#ai-dtc

What are you trying to do?

give data to ai, provide context and ask it to analyze the data

And what makes you think it's wrong?

when i look at the raw data to validate what its doing, its wrong

Kinda need more context here. What you're saying is super vague. And without knowing your exact goal, I'm not sure you're going to get much help.

i think what im doing is pretty advanced and expecting to receive information that i had two phd level data scientists do. sure it took them longer, but ai is not that good at data analysis without bumbling things quite often. i was hoping there’s some obscure ai out there that’s actually great for data/marketing/business analyses.

its not a singular thing or a simple ask like “did meta spend cause the first time customer revenue to go up?” - those it can do. its a very large ask of multiple data models to be run against each other to deliver a deep analyses and spit out trends and patterns

What AI is that?

Have you tried a college degree?

@Mark H - current AI chat tools are all large language models, which means that they have been trained on large amounts of text. But if you try to use the same thing for analyzing data, LLMs just won't cut it (it fails at basic things like addition and averages sometimes). At least for now. Some tools like rows.com claim to have reached 85% accuracy, but in my opinion, data is either 100% correct or just plain wrong.

You might be able to use it to answer questions where it just needs to pull a data point from the CSV, but not run any operations on it. For example, if you wanted it to tell you the number of sales on a certain date, both of which are on your CSV, then it can pull it out for you (on a good day).

I created this for an AI for retail article I wrote.

Did you try Shopify Sidekick? I have used it for some stores and shows pretty good results

Don't even need CSVs