AI Demo Jam is a new event series organized by Dnipro VC together with Data Phoenix. On March 20, six AI startups presented their products live during the first AI Demo Jam. In this video, you'll see the demo from CarbonCopies AI.
CarbonCopies creates AI-driven user replicas to optimize product flows and enhance website validation.
Transcript
Isabella: Hi everyone! My name is Isabella and I'm the CEO co-founder of Carbon Copies AI. So we help companies optimize their product flows and user flows by creating kind of AI twins of their human users. So the way that you sell to or the way that you build for a software engineer in San Jose or in Sunnyvale is going to be different from how you create for a Brazilian working mom, right? Different types of payment method, different consumer regulations, different consumer behavior and preferences. And yeah, I'm not going to spend too much time. I hope I gave enough credits to the demo god, so let's see. So in this case, I took a train down from San Francisco, and I want to make sure that I can get back to San Francisco again with Caltrain. So maybe we'll do a demo with Caltrain. So we imagine that we are a team working on the Caltrain products. So I can input the Caltrain website. Is it too small? No, okay. So what we do is that we probe the website, so understanding what the website is for, who are some of the user segments that might use it, and what are some of the user goals that they might want to do while visiting this website. Okay, so we generate a few user goals, and in this case, plan a trip, check service alerts, or look at fare. So let's do plan a trip. And say I am a commuter. Okay, let's change this. So I name all my AI copies Isabel-ish, because it's like Isabel-ish. So with the user goal and persona that we have chosen, what Carbon Copy does is it kind of maps out the different user flows. So one more, like lower detail on like each of the action that Isabellish might want to do. So let's say I want to, okay, plan a trip from Mountain View to San Francisco tomorrow morning. Sure, I can do that. Or maybe I'll do it tonight so that I don't have to… Tonight, March 20th. Okay, trip. And then we'll also check for parameters. Sometimes a lot of these flows are behind some sort of login. I'm sure this one is fine. Has it been five minutes yet?
Dmytro: Yeah.
Isabella: Oh, really? Oh, okay. Well, I'll maybe show the results quickly. So we kind of go through the website and be able to look at how the agent is clicking through it. We have the Persona insights, like what they observed and thought, as well as the UX optimization for it. We can also fix it with UI. Currently, we integrate with Livable to fix it. So let's say to make it more readable, to make it better understood. So that's that. Thank you. Apologies.
Dmytro: Thank you. Questions?
Audience: Yeah, I don't get it. I'm sorry. I know you were kind of… It's a beautiful design, and I'm fascinated by your user story you started with, the Caltrain story. And I couldn't connect that to what you had to sort of rush through. Basically, tell us what it is.
Isabella: Yeah, thank you. So after we ran it through, what the Carbon Copy does is that it will go to the actual Caltrain website, and it will click through. Think about, like, computer use, right? So it will click through the Caltrain website and think how to kind of achieve the goal of, you know, I want to get home by today. So what are some of the next available train? So we'll look through that. And if you guys are familiar with Caltrain, this is the website with like really messy. And it's funny because, so I didn't make this up. The Isabellish agent did say that. I see the Caltrain website. It looks pretty dated straight from the 90s. Like I didn't make this up, which is true. So what we do is we have a list of best practices that we train the LLM on and it helps to suggest certain improvements and how to help Isabellish kind of better achieve her goal. So that's kind of the…
Audience: I thought when you started, you were saying this is a tool for being able to create a bunch of hypothetical personas, like user and then test a website with them. So you don't have to get real people that test the website. Is that still what you're talking about here?
Isabella: That is correct. So this is the outcome of the test and the improvement that you recommend.
Audience: From Isabella's perspective.
Isabella: Yes, that is correct.
Dmytro: Thank you.
Isabella: Yeah, thank you. So you can scan the QR code.
Comments