Latest / Talk Commerce / Email Marketing in Agentic Commerce with Rahul Talari of Relvino
Transcript
- Brent Peterson (00:06) Welcome to this special Agentic Commerce edition of Talk Commerce. Today I have Rahul Talari. He is the CEO, co-founder of Relvino. Rahul, go ahead, do an introduction for yourself. Tell us your day-to-day role and one of your passions in life. Rahul Talari (00:21) Awesome. Hey th Brent, thank you for having me here. I'm Rahul Talari, CEO of Relvino. I was previously second ML engineer at CoreWeave, the AI cloud behind OpenAI, Meta, and Nvidia, the only AI company to go public, where I was involved in building early models for OpenAI like the ChatGPT 3 and 4. And my day-to-day involves working with brands, helping them automate marketing with our platform versus the legacy rule-based tools. and that's what Relvino that does is it helps brands market to shoppers. Brent Peterson (00:52) That's awesome. How about passions outside of work? Rahul Talari (00:54) Absolutely. Yeah. So I have a I have a lot of passions. one of my core passions is jujitsu. I I do it a lot. helps me think strategically since it's kind of like chess, but for physical, where you actually have to use your body to to manipulate and get to a dominant position. I also run triathlons, recently started that. not as much as you, Brent, not 34 marathons, but getting there and Yeah, other than that spending time with my son, and my family and and anything outdoors. Brent Peterson (01:23) That that's great. And when we have we'll we'll have a longer talk about triathlon and Jiu Jitsu. My son is a big into jujitsu, but before we do that, I forgot to tell you, I do this thing called the Free Joke Project. I'm just gonna tell you a joke. You give me a rating eight through thirteen. It'll take about fifteen seconds. So here we go. Rahul Talari (01:41) Sure, sounds good. Brent Peterson (01:42) While most puns make me feel numb, math puns make me feel number. Rahul Talari (01:46) I I I think I get it. Brent Peterson (01:48) You probably have to read it 'cause number has a be num has a B in it. Number has a B. I know. Yeah. It was a stretch. It was a math stretch. Sorry about that. Rahul Talari (01:57) Th that was that was that was intellectually engaging, so so I liked it. I I'd rated Brent Peterson (02:00) Yeah. Rahul Talari (02:03) I'd rated four out of five. I I just feel like Brent Peterson (02:05) All right, thank you. Rahul Talari (02:06) maybe maybe I need to dwell on it a little bit more. Brent Peterson (02:09) Yeah, perfect. yeah, if you're on again, we'll have we'll we'll have a new joke for you. You c you'll be more ready for my non-intellectually engaging puns and and jokes. So Rahul, this is this is meant to be just a quick talk about agentic you're obviously involved in agenc the time in a green room. You talked about all the agents you have running. but I think I I do wanna talk specifically about email and where people can really go wrong a when they're using agents on their email and where they can go right. Tell us, give us both sides of the coin, where people really can dig in and do the right things and where people have gone off the rails to just do the wrong thing. Rahul Talari (02:46) Yes, that's great. So let's start with what's wrong before we go to what's right. So the way today marketing works and how teams do marketing, especially with email and SMS, on on tools like Klaviyo and Attentive is you send a cart abandonment email once someone adds the cart and abandons it two hours later. You send them an SMS ten hours later. And you do this rule-based approach of assuming when people are going to interact with your website. based on triggers that you set manually, where your marketing team and data silos that you're analyzing. So these signals that you're capturing are not actually based off of user behavior. They're based off of assumptions based on market research as to when someone would come back and shop on your website. This is what is called the spray and pray approach, which is absolutely the legacy way of doing it and how incumbents do it. And the right way to do it is to really understand the shopper and what they want as a person. So as you, Brent, and not a cohort that looks like you in Massachusetts who has a family and is married or in Hawaii. But you, Brent, as a person, how you shop. And and what Relvino does is it helps understand the shopper and engage them at the right time with the right content through the right channel. So we're not spraying and praying a thousand emails to a thousand shoppers who've added to cart abandoned. We're only sending 50 emails By identifying the strongest intent and providing them the right content at the right time through the right channel for them to actually engage and convert. So this is what is kind of the state of the art of these rule-based systems that send emails or SMS based on triggers versus actually modeling user behavior. What's going right is to be able to allow for the brand to to set their guardrails and to to allow for the tone, the voice, the strategy. the products and the the kind of image that they want to put in front of their audience is what is going right. That will always stay in the human's control and the brand's control versus the volume, the repetition and and the automated way of marketing based on user behavior is is what is something that we're trying to solve for right now. Brent Peterson (04:59) Yeah, so the the difference in that rule based is is the if then statements, right? And the and the the the may the newer approach is going to be where we're using some agentic, some large language models or small language models to help better understand the intent of the user. Is that correct? Rahul Talari (05:17) That's right. So we don't have flows, no more segments. Every single shopper has their own marketing agent based off of sparse data we collect on the website as to how they click, browse, abandon, or convert, offsite data as to how they click, open, and purchase through email SMS, and any other data that we can get from the CRM about who this user is to market for them as a person. Brent Peterson (05:40) I think one of the things that we've been hearing since email started is that email is dead. But every year we hear these new things that are happening in email, and it just continues to be the best or the mo the best channel that we can most frequently use, and we don't have to worry about a platform going away, right? So maybe and also it's most it's really easy to to make it. personal for the person getting that email. So SMS and email probably be in that same kind of category. How do you rank those two and then against like all the other social media stuff that's out there that people are also pushing? Rahul Talari (06:18) Great question. So the way we're thinking about solving this problem is to focus on engagement and retention first, which is lifecycle marketing when a user lands on the website to when he purchases or abandons, to when he comes, he or she comes back on the website. And how how we're thinking about this problem is solve engagement and retention and model user behavior as they're actually shopping to be able to then plug into discovery, which is the social media or ads where they discover a product. So we're flipping the coin on its head to be able to go engagement, retention, and then discovery. And I think that's an important aspect of being able to do email and SMS correctly because once we model and understand what users are actually doing on the website and off the website, we can then go into discovery. And it's worked really well for us. We're usually seeing 80% fewer sends than current incumbents like Klaviyo and Attentive. On average, two to six X revenue uplift. on brands that are doing 100 to 500 million in revenue, open rates of 35% versus the 26.6% industry average, and in general about $7.64 in revenue per email versus $1.07 that we've seen across the the incumbent stack. So it's really working when you're actually targeting people the right way and landing in their inbox versus in their promotions or spam folder where they click unsubscribe or mark it spam and you're losing delivery deliverability and your audience that way. So we're very cautious about how we actually send when we send and what we send to be able to alleviate the problem of email is dead or SMS is dead with with renewing this process to actually delight customers. Brent Peterson (07:58) I wanna close out the segment just talking about AI slop and and and non-human intervention and the messaging they're using. How important is your message that's going out and that that message is relevant and that that doesn't sound like an AI bot. Rahul Talari (08:14) That's a great question and something that we're very careful, especially because we have to adhere to brand guidelines. What we do is we have a QA layer that's non-AI and written purely software based and allows for all these AI systems that we've built to be passed through this layer to verify all the work that the AI did. And we score based on a certain threshold as to whether it passes or fails. And if it fails, we go back and do the entire AI loop again. Within the constraints of your guardrails. So we set global guardrails for your brand, local guardrails for your campaigns and flows, and we strictly adhere to these guardrails based on a software layer we've written that's that's QA-based and non-AI. And I think this is very important to be able to adhere to these.