Two speakers explore how LLMs fit different marketplace stages. Early-stage shops like Fixy use them to automate returns on Shopify. At massive scale, proprietary data builds AI flywheels that expand categories. Service marketplaces may see faster LLM payoff than product ones.
You did mention, as far as the point is, you know, unique datasets, right? And so if we think about, you know, data in this kind of sense of scale, you know, with marketplaces, and we think about, you know, opportunities where LLMs can be kind of applied to them, do we think that, like, if we look, you know, say, look back, you know, a few years from now, that it's going to be like a specific kind of stage of marketplaces where it's applied, you know, more effectively, or?
Yeah. So let's say that you have an e-commerce business. You know, if someone wants to exchange a t-shirt right now, you need to have someone in the business that will basically man that. And, you know, manage the— go into order history, look up the stock, do a return, and then, you know, reply back to the person. Now there's a company called Fixy that will basically— it's LLMs wrapped around, you know, Shopify endpoints that can do that. So what that means is that if you're trying to solve a cold start problem, like you can create some scale and automation for yourself that your— let's say your peers or predecessors didn't have, which just allows you to move faster. So that's how I would think about LLMs for that stage of company. For a massive scale, I think at that point, when much more about like, how do you create, like, I could share this great example where they're, you know, they're post-Series C, they have a lot of money in the bank. It's all about, we have these unique datasets, how do we leverage those and create and be the go-to place? Like, how do we make the flywheel spin faster and conquer more categories using the LLMs where maybe in the future our business becomes a pure-play AI business just based off of the fact that we've collated the the collective crystallized intelligence of all the agents right now. So then in the middle, it's a little bit messier. And this is where I think a lot of nuance matters in terms of what's the nature of your business, what's the core value that your users get from it. And so I'm not— I'll sort of demur, but I'm happy to, like, if there's any founder here that's at that middle stage, happy to chat one-on-one or chat here about your specific example, just because I don't want to offer like a generalized solution that's not that useful.
So we did cover, you know, stages, right? And so kind of the cold start problem, then later some sense of kind of scale. So what about marketplace types, right? So like products and goods versus say services or kind of spaces. Do you see any that are uniquely positioned to leverage it more now in the kind of short term with more initial adoption?
Companies that are focused on in a few different sectors, but focused on some sort of coaching or there's a human in the loop where it's executive coaching, product management coaching, health coaching, weight loss coaching. I'm seeing these companies leverage with some fine-tuning of prompt engineering, you know, GPT-3.5 or into creating like 5 to 7x efficiency when it comes to the supply side of their product and offering. So that's super exciting to see from a few months of experimentation, let's say a quarter of experimentation to Starting to see those efficiencies really allows the business to scale up. So that's exciting from a service perspective.
