Bringing a Generative AI Product to Market: RJ Talyor, Founder CEO, Backstroke and Pattern89 (Part 5)

Sramana Mitra: Very good. So, what is the genesis of the new company?

RJ Talyor: Well, when we were purchased, I was playing around with the early versions of the generative text content. With my history in email marketing combined with my background as an English major [it’s really unique in the tech space], I went back to what if we could create unique and performant content that drives actual performance on the email side? So, those are two important adjectives, unique and performant.

We started creating a giant dataset, and over the last two and a half years, I’ve signed up for thousands and thousands of email marketing programs across the globe and done surveys out to panels of consumers. So, we have this longitudinal data around what consumers prefer, combined with what is the popular thing that is being done by global retailers. The idea is, what if we could help marketers create unique and performant content?

GPT, of course, exploded about a year ago, which further accelerated the adoption. It was a great moment for us. We went out and pitched this idea for Backstroke to five VCs, all of whom except for one I had worked with before. All five of them were interested.

Sramana Mitra: Are these local VCs?

RJ Talyor: Yes, they are Midwest-focused.

Sramana Mitra: So how much did you raise?

RJ Talyor: $2 million.

Sramana Mitra: Now, GPT and Unique are hard to go together. So, explain how does GPT help you be unique? GPT is available to everybody and their mother.

RJ Talyor: Yeah, so what our system does is it’s subscribed to thousands of brands, and each of those brands has a certain target and sells certain things. So, a clothing retailer might be a luxury midpoint or a discount retailer sell jeans, leggings, shirt, or lingerie. Each of those brands has a unique voice. They have certain brand terms that they use or that they avoid.

So, what we’ve done is, we’ve created a fine-tuned model that identifies at the aggregate level for industry as well as for price point. So we start with industry, then we go down to price point, then we go for the types of products, and then we go to the brand voice. So, we can get very unique and on-brand as we further fine-tune the content.

The second piece is performant. You can create something that’s really unique but so weird that nobody’s interested. The combination of unique or differentiated with performance prediction is what marketers truly want because they want to sound unique but also drive purchases. So our fine-tuning has taken into account both of those so that you can create content that is performant as well as unique.

Sramana Mitra: Is it built on top of ChatGPT?

RJ Talyor: Yes, it’s built on top of the GPT 4 model, but we built it so that it can be lifted and shifted to whatever model might be going forward. We’ll apply that as appropriate, but right now, we’re seeing outstanding results in terms of the performance of messages that use the subject lines that are generated by our system.

So, we’ve got a vision for generating any and all content on the messaging side, specifically CRM – email, text messaging, push messaging, etc. We’re starting with email subject lines, and the good news is that it works, and some of those early customers who are both free validators as well as now paid are seeing outstanding results with our software.

Sramana Mitra: And did you focus on apparel as the first category?

RJ Talyor: We have three focus area: Apparel and Fashion, Pets, and Beauty. So those are three distinct areas and distinct models with pretty aggressively different approaches. We started with Apparel and Fashion, and then we extended out to Pets, and then third was Beauty. We’re going to add other channels or other industries here shortly.

This segment is part 5 in the series : Bringing a Generative AI Product to Market: RJ Talyor, Founder CEO, Backstroke and Pattern89
1 2 3 4 5 6 7