Everyone says AI needs huge datasets but I'm proving them wrong with a local business campaign

I trained a custom chatbot on just 47 customer emails from my landscaping company and it generated more qualified leads last quarter than the $500 monthly ad spend ever did. Anyone else find success with small, messy data instead of cleaning everything first?

2 comments

2 Comments

the_derek1mo ago

Man, YES! That's the stuff right there. People get so hung up on perfect datasets and millions of samples that they forget real businesses run on real messy conversations. Your 47 emails are probably WAY more valuable than some giant generic dataset because they're YOUR customers saying YOUR specific pain points. I've seen the same thing with small local shops - a tiny batch of honest feedback beats a mountain of polished data every time. It's like people forget that every big AI started as a little experiment with whatever junk you had lying around. Keep going with that dirty data, it's gold.

elizabeth_williams1mo ago

Wait, only 47? That's insane ROI for basically zero data cleanup. Makes me wonder how many emails they could have started with.