Your Shopify Support Data Is Worth Millions

Jay Myers: Most brands I talk to, they think ai, customer service is about reducing headcount. How can we put in AI to have less people on our support team less cost. But the brands that are actually winning with ai, they're using it in a different way. They're using it as real time intelligence on their products. How their customers talk about it. It's actually becoming a bit of a intelligence layer that's changing how they run their company. Not just product decisions, but inventory, marketing, copy. It's all driven by what the AI is hearing in support conversations, and this is something that I think not enough people are talking about, which is why I wanted to bring my guest on today.

He's the co-founder and CEO of Sienna, which is. Some people say AI support, but it is so much more than that, which is what I'm excited to dive into. 'cause I think, I know probably 95 or more percent of the brands I talk to are not using AI as much as they could be. So Andre, thank you so much for being here.

First of all, you wanna give a little quick background on you and Sienna and we'll jump right into it.

Andrei Negrau: I'm excited to be here. Thanks for the invite and super excited to dig in.

Jay Myers: So we had Lisa from Sienna back on the podcast. About a year and a half ago. And so a lot of people listening right now may not have heard that episode. It was actually episode number one and it was awesome. It was I don't have the rankings up, but it was one of our top episodes probably because it's the first one, I think a lot of people go back and they listen to the first episode in the series, but it just feels like the. World of AI is changing so fast and already I remember things we talked about in that episode, it feels like a lifetime ago, and I know so much has changed with Siena since then, but what's the biggest shift you've seen in how merchants think about ai and in the customer experience from what, a year and a half ago till now?

Andrei Negrau: Year and a half ago. Sounds like a

Jay Myers: Year to two years ago,

Andrei Negrau: It just sounds like a decade ago.

Jay Myers: I.

Andrei Negrau: So it's really interesting what we see happening. It's almost like a pretty, pretty big division in terms of companies and, you know, you've probably heard this term, AI native. Let's just use AI native for the purpose of this podcast.

AI Native doesn't just mean they use the usage of tools, it's also what we see is it's actually a philosophy of running the company.

So in the last year and a half, I think the biggest shift happened on a more philosophical slash structural slash operational level, where it's not just like a cute project or a, Hey, let's build a chat bot, let's put a chat bot on our side.

But really, we see brands irrespective of their size, become a lot more intentional about the way they think about AI in, in the general sense of. Deploy AI in the company, and that's been incredibly exciting to see a year and a half ago. You can imagine this diffusion, I think most recently people call this like the diffusion layer.

How do you embed AI in different parts of your business? Was quite immature. Again, very, you know, very limited. Number of use cases, you can think about content creation. It was already quite there or support automation. But now we're seeing more as a, can we run as a company or how can we run as a company on ai?

And then by asking that question, then as a Shopify brand, as a brand, as a company in general, it doesn't really matter what you're doing as long as you extended digital realm. You start thinking about different avenues of over or like different ways you can embed AI in your workflows. And I think that's the biggest shift.

Because the tools are here, like we've seen a massive wave of tools being created. The models get better, although the model if you're like following you know, the latest models, the progress I wanna say stopped, but it's definitely a lot slower than it was two years ago or and a half ago, where you have like new models almost every month now.

It's yes, there's models out there, but we have. Today's models are already quite good for most tasks. So it's just really a question of, in a company who assumes ownership over AI in general and then how do you create the culture where people are excited about ai? So I was I'm always excited about seeing this firsthand by talking to our customers, by talking to brands, by seeing what's on their mind.

But it's really exciting because the folks that come to us and wanna partner with us, they know. They have a pretty good idea of what they wanna do. And they're treating, obviously, they're treating AI as a real thing, not just like an experiment. So that that's changed quite a lot in the last year and a half, like just philosophy about ai.

Jay Myers: Yeah. What percentage of stores would you say that you talk to are still treating AI as a cost cutting tool versus a growth tool? Like where's the breakdown in that?

Andrei Negrau: I would say that in the beginning phases of AI for customer experience, most companies see it as a cost cutting tool or as in, maybe it's not cost per se, but it's an efficiency gain because what we also saw, we have some pretty fun story with brands that joined and implemented Siena, let's say a year ago when they were doing.

25 million a year. You know, just on that upward stricter. And now they scaled to hundreds of millions in the span of a year. So they didn't really come to us to cut costs per se, but they came to us because they wanted to not having to grow their teams and their processes and their overhead linearly.

So we see a lot of folks do that, and they do it really well. And I think that's one of those misconceptions where hey, you know. Customer service, ai, it's like replacing where you know, it's cutting jobs. I think that, you know, thankfully the brands that we work with are growing really fast.

So we rarely see someone just go from, you know, you know, you pick a number to cut 90% of their support. It's really the opposite, which with their support team stays, but they stay intact because their business grows. So it's more of an offset mechanism that allows you to just move so much faster because now you have AI that Siena running, you know, 60, 70, 80% of your support interactions or even more.

Jay Myers: Yeah, so tell me about what are some of the. The best brands, how are they using AI with customer support? Not just for answering tickets, but what are some of the exciting ways they're using it to grow their company and improve operations or improve their marketing, their product? Like any other besides, okay, answering tickets is one thing, but what are some other ways they're using it?

Andrei Negrau: Yeah. So there's, in the ecosystem of agents that we're building, there's the agent that resolves customer conversations. That's our flagship customer service agent. And we've been running this agent for more than three years now. We've been, I think, the first company to build an agent using large language model.

So that's one agent that already works really well. It's scalable, works in the enterprise. And the other agent that we're seeing now, brands really leverage. And just before this podcast, I was here in this room with with one of our brands and we're brainstorming ideas. How can we use the data that CNA already has with this new agent?

So this other agent is part of what we call CNA intelligence, which is our up and. Platform. If you think about cloud code for cx, that's what Aska is. So a lot of novel ideas, or a lot of new use cases are spinning out by using this additional, like the second agent of our called Aska. So just for those that for those of you that are new to Siena and you know, hearing it for the first time it's basically an agent that sits on top of your.

Starting with your support data. So all your conversation data, your reviews data, your social media data, your Shopify data. And now we're bringing more data sources like subscription data. So we really are building this brain that's all powered by, of course, the best agent, the latest that the latest model.

So the question the question that we've been just discussing an hour ago with with this brand is how, you know, what are some of the things that we can do to retain our customer? SNA has already that, like we already know how many customers reach out to cancel a subscription. For example, Sienna knows that you can easily ask Sienna, Hey, give me a list of all the customers that reached out in the last seven days to you know, to cancel their subscription.

And then

Jay Myers: Through a support ticket.

Andrei Negrau: through a support conversation. Yeah, and what's really interesting is of course. There's very various different agents doing various different tasks in, you know, today's day and age. You can even ask probably cloud code to connect it with your Zendesk or to connect it with your help desk.

But the challenge there, or the thing that we solved with this product is we've built we've built a pipeline that allows us to essentially filter or thread the signal out of the noise in customer service data. It's one of those things that it's really messy. If you think about any given take any brand, most of the conversations are either repetitive or there's just a lot of noise that bear the signal.

So what we've been able to do is we've built a whole pipeline that allows us to filter and only give you the real signal that matters based on what you care at that specific point in time. For example. Is really good at finding the new in the haystack so you can actually ask it, Hey, can you go and check out who are this?

Cus you know, gimme a list of customers that reached out to cancel their subscription. But it, you can actually get quite creative because then you can say, but categorize them based on the reasons. So imagine you can do anything once that data is there, you can ask it to do anything. And one of, one of the cool things that.

We've been discussing about just recently was the idea of creating win back campaigns that are one-to-one personalized for that customer. So think about the old world is something, you know, someone cancels. Subscription. And if we're talking about a bigger brand, we're talking about probably hundreds or even thousands of people canceling subscriptions every given week because it's just like a lot of like new customers turning and so on and so forth.

And in the old world that triggers some Klaviyo flow or you know, that triggers some sort of an SMS flow and then, but it's something like, hi j, I saw you canceled your subscription. I would like to offer you a discount if wanna, based on some sort of a, you know, variable, high first name that's. So what we see now, customers, and this is not like an this is, we see other brands do this with Siena.

You can literally ask Sienna, Hey, can you find those customers? Look into their entire conversation history with each individual customer, look into their Shopify orders, look into their subscription, which we're building now subscription integration and create. Create a win back in pain. And it's gonna take a little bit of time.

Jay Myers: Mm-hmm.

Andrei Negrau: Think about it. In a non sienna world, if you don't have Sienna, that's probably gonna take you. If you have hundred, you have manually go, you know, export customers, go see what they said. Probably more like days than you know.

Jay Myers: And so because it has access to the ticket. And the reason they canceled, you can tailor a winback campaign, like if they thought it was too expensive or if they said in the email, I don't need this, whatever product anymore. Maybe it offers a different one. Is that

the idea? Like it's tailored by

Andrei Negrau: It's hundred percent and there's no limits in terms of what you can do because we're.

Let's say mental you know, mental challenges, like mental tasks and they can like compute and do the good work. But yeah, overall you'll see some pretty incredible things. Like I was just reading some of those examples of winback campaigns. It, by the way, it, in it, it also automatically knows the channel that this conversation happened over.

So it's gonna automatically created for sms. So it's gonna be short. It even adds without me even prompting it stop, you know, stop to, you know, stop to cancel. That that legal language based

Jay Myers: Yeah. Yeah.

Andrei Negrau: previous conversations and so on and so forth. This all looks like a human has written it.

Once you read it, it's like there's no way we can go back at like the old of like high first name that I think that world is going away and. In the realm of CX data. I feel like this whole idea of, oh, CX is a growth channel, or CX data is, or like customer data is the most important data that, yeah, that's kind of like an aphorism that existed for the past 10 years, since even when I was running my own brands many years ago, it was like CX data is the most important data, but I think we're at the stage where this really is a reality, like you can do things with it.

What I love doing is really sitting down with brands and really sitting down with. With folks that are already so good at understanding how the data fits with an lm and it's all a matter of prompting and it's all a matter of creating these processes. So that's what I'm spending a lot of time.

Most recently it's okay, we've built an agent that can do support really well and that is something that still needs a lot of like work to make it even better over time. But this new generation, what I think, you know, what do we call it? CX marketing or marketing in general, or just growth that's gonna happen as a result of connecting all those things together?

I think it's gonna be a wave that won't happen overnight. It's gonna start slow, but once you unlock that, once you let the genie out, it's gonna be impossible to go back at you know, very basic flows.

Jay Myers: Right. So this is this live yet right now? Or this is something you're. Experimenting with brands like

Andrei Negrau: On when, depending on when the podcast will

Jay Myers: this goes out, Yes.

Andrei Negrau: We are we're live. We're live with some, with, we're live with customers as we speak. It's

Jay Myers: Okay. And so

this is, primarily, this is it also a tool that merchants can ask questions to about their customers? Or is it

you

Andrei Negrau: Yeah.

Jay Myers: give it a task and it does it like. Is it merchant facing or is it customer facing, I guess is primarily the question?

Andrei Negrau: So Sienna Sienna, the support agent is customer

Jay Myers: Correct. Yes.

Andrei Negrau: That is available across all the channels, including voice, including social media, SMS. This particular tool that I was mentioning, we call it Ask Sienna because you're asking Sienna something that is an internal facing. So that's something that you as a operator in any department can interact with Siena, and it's gonna answer questions, it's gonna create reports, it's gonna generate charts.

It's really, yeah. The best way that I explain it to someone who's new, it's like cloud code, but for cx

Jay Myers: Yeah.

Andrei Negrau: and even more than that.

Jay Myers: And so for a Shopify brand, the reason why this is probably more valuable is it has, like Shopify's got its sidekick, but. That doesn't have access to customer interactions, like that would be the a big data layer that's missing.

Is that Correct. Yeah.

Andrei Negrau: Yes.

Jay Myers: I know Sidekick is quite limited. You can just ask it queries about your data and stuff, but it doesn't like the example you just gave, create this Winback campaign. I don't even, I don't think it would do that. Where does Sidekick play in this? Is that is it's starting to come up more and more with merchants is do you draw a line anywhere of okay, this is a good job for Sidekick, this is a good job for Sienna.

Andrei Negrau: The best way to think about jobs to be done and which AI to use for what it's, I think. You can reduce it to the, what context does the tool have access to?

Jay Myers: Right.

Andrei Negrau: The underlying agent or the underlying model? Probably most of companies producing or building, these agents today are using a combination of child GBT and Gemini.

Maybe some of them use some like open source model so the underlying models are, I would say, roughly the same intelligence level. Sure. There's different frameworks of how to construct these agents, what tools to give them, and so on and so forth. But roughly, they all work in similar fashions and similar tools.

The only thing that really matters is the context that it has access to.

Jay Myers: Right.

Andrei Negrau: So I think Shopify's STIC is incredibly useful for anything you need in Shopify. If you need something in Shopify, that's your go to. I think it would be unwise for a company like Cena to build something that directly. Or tries to replace what static is doing.

It's sure, at some point we could build some tools that allows you to maybe make changes to your store and do something a little bit more in that like web app kind of experience. But for now we're looking for, we really started ask as a way to first solve the hardest problem, which is this like conversation data piece.

That is by far the data that is in abundance. There's a lot of it, but it's really hard to find. Find a way to make the, to make it useful without blowing up your tokens. Again, you could export technically, you know, you could hook up cloud code to your Zendesk or your, you know, your help desk, export millions of data points and then ask it to run something.

But that could probably, that will probably cost you a lot of money and you don't really know what's gonna give you back. Truthfully, it took us quite a bit of time from the whole idea of Ask Cena to actually having it. See in production, you know, providing useful responses and inaccurate very important, accurate responses to the users that it interacts with.

And it's a lot of, it's a lot of heavy work that happens behind the scenes that you don't know about. So for someone who's, for someone who's thinking about, Hey, how can I get smarter by understanding what customers are saying? I think we're the only company, or CNI is the only product that can do this today.

Connected ecosystem continuously. And again, this is a very much new product and it's up, but the power that we're seeing with the first version it's pretty incredible.

Jay Myers: That's amazing. I mean, it is, as far as I know, it's the best product for it as well too. I don't think there's anything else really. I know different tools, like Zendesk has their own version of some insights, but I we use Zendesk, but it's very primitive.

Andrei Negrau: Not for long.

Jay Myers: what it No, I know. Actually, yeah, no, I would, it's in conversation as we speak, but okay.

So if with Ask Sienna, if you created, let's talk about like the, say the Winback campaign. How does Sienna do some of the, like messaging? Does it, you have to have it connected to gorgeous or Zendesk or Intercom or some tool, or what does that look like?

Andrei Negrau: So we're building some of those channels natively. You'll see some channels being available inside Siena in, in the coming months. And for those that. Are already using a help desk, we can already reach out via the existing preexisting channels. Yeah, you're gorgeous. Or Zendesk. So it's gonna be, it's gonna be a model where we're designing first and foremost.

You know, compliance is important. So if you're talking about something that's more, becomes more of a marketing thing, then that's a separate conversation. That's a separate. Let's say motion, CX motion that we're looking to design. And then the channels will have to be, of course will have to look at the marketing consent and all those things.

Jay Myers: I guess also like Klaviyo or Omnis Send or other tools like that as well too. Potentially

Andrei Negrau: Yes.

Jay Myers: depending.

Andrei Negrau: so from the beginning, our philosophy as a company was leverage the tools that already exist. So if we can generate this campaign and then push it to klaviyo's, or omni, or attentive, or you name it, then that's gonna be our preferred way, because that's the fastest time to value. Our goal is not to rebuild something that exists for many years unless we can do it better.

The only caveat here. Sometimes these tools I, you know, I don't wanna call them legacy tools, but let's call them pre AI tools. Some of these pre AI tools, they may have actual physical or architectural limits that does not allow an, or an AI first tool like Siena to run its course. And you know, do these, let's say the one to one-to-one personalization.

So as long as we can accomplish. The delivery through a third party. We're gonna do that. Like we would much rather connect, hook up to your stack and do that. When we hit a limit, we're gonna be like finding ways to build this natively just

Jay Myers: gotcha. That makes sense. What's something, I mean, you talked about like this idea that su, the support inbox is basically a focus group for learning about your product and your brand. What's maybe the most surprising business decision you've seen a brand make because of some CX interactions?

Andrei Negrau: There's there's quite a few of them. And because we work so closely with customers, we get to hear about them. One of, one of those decisions that I always get back to, it's quite fun. It's a CPG brand that got on, got onto even before CNI intelligence we built our. Like early concepts and early prototypes.

So they've been an early adopter from the beginning of all the various iterations of our product. At some point they made this decision to build, so they sell CPG products and the products are not super expensive, but they're not cheap either. What they saw through Siena by looking at the support conversation data is that customers were either abandoning their carts.

Churning or requesting actually like making requests for smaller samples of their product. So they were just selling regular, you know, regular size products that it wasn't like it's prohibitive, but also people just wanna try. If, especially if it's a CPG product, you just wanna try it, you just wanna sample it, you maybe don't wanna pay full price.

And that little nugget right there led to them

A full, basically they made this product decision to build for each one of their products. A small, like a, think about a simple size version of their product. And they generated ever, this was happening a few months ago, back in 2025, and they generated millions

Jay Myers: Wow,

Andrei Negrau: samples.

That's sample products, not including the people that upgraded

Jay Myers: amazing.

Andrei Negrau: the full product. And this is all because they saw in the data that, hey, people are. Outright asking us, Hey, can I try it in a sample? And they, what's amazing about Siena, it actually is able to even quantify to some extent it, okay, so there's you know, 500 people ask this in the last 90 days.

What if their LTV is this much over, you know, this time? Okay, this is the investment. And so yeah that's a real example that happened by looking inward at what customers are saying.

Jay Myers: So what does that look like? Is there a dashboard of insights or do you ask Siena for details? Do you ask it questions or is does it find some of this stuff on its own and surface it, or what's that? Experience look like for a merchant.

Andrei Negrau: So we have a few options to interact with your data. One of them is your traditional dashboard. So we do have things like trend explorer, so you can see different trends. For example, you're able to take a look at top. We do on the backend automatic topic and subtopic clustering. So in the, you know, in the old world, you'd have human agents label a ticket.

This is a cancellation request, and the reason is this. You know, you'd have to spend time and mental energy doing that. Plus it's not accurate most of the times. So we're doing all that. That data is available for anyone to see in a very simple dashboard. So you can see the different trends, you can actually see resolution rates, you can see what are the topics that are being automated the most, what are the topics that are not being automated.

So it's a hybrid between your full CX picture plus

Jay Myers: So basically like themes, like if there's, people have like breakage in a product, bad shipping or questions about order, timing or whatever. You can see where. The majority of the

tickets are, and then potentially dig into that and more.

Andrei Negrau: exactly. Yeah.

Jay Myers: Is there a specific theme? Sorry, go ahead.

Andrei Negrau: Yeah, we do it at, we do it at two levels of fidelity. You do it in a like bigger theme, as you said, and then you can actually go for each team, we do sub topics, so you can go even deeper. So let's say you see a spike in refund requests. So refund requests could be a theme or a topic.

And then you know, hey this week or this month, you're seeing a spike. Then you can click on that double click and then see exactly why, what's the reason. Because Sienna was smart enough to figure out there's multiple reasons or there's multiple.

Jay Myers: Yeah. Is there any FI things that merchants find using this topic explorer that is always shocking, like you might think, I don't know. Sometimes we have blind spots as merchants and we overlook things or we get. Blindness to the same questions. Like we talk about this all the time, like when you're in the weeds of something you get, you don't see it through a new set of eyes.

Right. Is there any like themes that consistently surprise merchants when they go into them?

Andrei Negrau: Yes there's always this, there's always this reaction when we look at the themes and, you know, you can look at it for seven days, 30 days. You can go, you know, more than that. And then they look and see some sort of a spike or they see something. Sort of a trend. And let's say that we, let's say that we're doing this live, we're, you know, like we're looking at the data live with someone who works in cx.

It can be from a support agent to a VP of customer experience. And they're looking at this dashboard and sometimes we see these directions like. That makes sense. We knew about that, but there was never, the data was never there to support that argument. So it's almost like sometimes in a pre AI world, what happens is for CX teams, they see some behaviors, they see some trends, but they don't have the data to quantify it.

So they, no, they could be spending weeks or even months trying to prove something out. But no one is really gonna take weeks or months to actually track tickets, like tag them and put them in spreadsheets. And so what this gives them, it's probably like a in five seconds. Look at that. Like we knew that returns are increasing or we knew that returns requests have been increasing and now we can see why.

So we can go deeper. And in one click you can go from that dashboard to actually asking Sienna. Hey, can you tell me more? What's going on here? So that's always an aha moment where you can quantify something that you maybe felt, or your team has anecdotally reported to you hey, we're seeing more cases of this product being mentioned, or something like that.

There's no real way to track it. And when you look at it in that dashboard, you're like, okay, now it's real. We can, you know, we can share this report with our operations team or our product development team.

Jay Myers: So say I'm a supplements brand, average size doing five to 10 million a year or something. What's. What's a question I should be asking Siena of my data for? For sure. I should be asking what are like, or it doesn't have to be a supplements brand, but what is, can you give some advice on what the best, because often I, it's asking the right questions is the key, right?

Like you can have, data is not the problem anymore. It's so what are some questions that brands should be asking it?

Andrei Negrau: There's levels to this. So that's a great question. And there's levels, right? You can start off with pretty generic questions. One that I always like is take a look at my data. And of course we have ways to build prompts in a more robust way, but just riffing here, I always like the idea of, Hey, look at all my, look at all my happy customers.

So I would, I'll always do almost like a workflow base. So look at my happy customers. What do they love more about the. What they love, what do they love most about our products and why? So give me a summary. Then I will go into, Hey, what's some constructive feedback? So look into what are they not liking as much?

And then eventually I'll say, what are the business? What are some of the biggest business opportunities for us? And you know, you can replace business with what are some of the biggest product opportunities. So often, like we've seen brands use CN Intelligence as a way to do product development.

Thinking hey, okay, so we see these many customers asking about this. This is interesting. Oh, like we never thought that we could actually just do this. So anything that, you know. Think about it, you have access to this, all of a sudden it's a gold mine. It's like a oil. You, so you can, you know, you can dig for oil and oil comes out of it.

It, the question is, you know, how do you make this oil actually useful if it just sits there and you look at the oil, it doesn't really do much. So you have to make, you have to transform into energy. What kind of questions can you ask? What are some of these like high leverage questions that you can ask?

I would start with what are we doing well? What can we improve? What are, I also love this question. You know, often you, you grind hard, you work hard, and you kind of forget Hey, you have a, you have an amazing community that is behind you, and people love what you're doing. And this is, I think, quite challenging for many folks that are working in customer experience.

Customer experience, for the lack of a better word, sometimes equals just like people complaining. Not always, but. You reach out to support because you have a problem very, you know, very limited. I guess like the pie chart of people reaching out because they love the product is smaller. But I also love to, to show what's possible when you ask Ena to show, to give you reviews.

Hey, show me some positive reviews or another use case that I love. And it's really helpful. And you can, by the way, you can, what's something really cool about. This process, you can run it automatically. So you can just schedule it and run it once a day, once a week, once a month. So you don't have to manually always like type in what are customers saying?

You can just develop your prompt and then that can run on your own schedule. So one of them could be, hey, can you create a, can you create a a list of some of the best testimonials or reviews that my marketing team can use? And can you gimme also the angle that my marketing team. Can use when it to that so the cool thing about it is the system will know a few things.

So it will know what products they reference. It'll know the actual customer. It'll know the review itself because it looks at review. And by the way, when I say review, it can be a social media comment, it can be an email. It doesn't necessarily have to be a review channel. It can come from anywhere. And it also knows the brand itself.

So now we basically have given this AI a task to create. A campaign based on customer voice or customer customer data. And that's also something that I highly recommend anyone. Hey, what are my loyal people? You know, what's my loyal audience saying about our products? And it's really hard to have a bad day once you see that, Hey, we have a lot of customers that love our product.

Here's who they are, here's how much they spend. Even better. You can ask to create a personalized outreach campaign that focuses on something like, Hey, you're a lawyer. You're a loyal customer because X, Y, Z, and then you ask to create that. So the whole idea of hyperpersonalization, how. I call it then, you know, in the past it used to be personalization, but now you can truly have these concierge type of experiences that you create for your most loyal customers because you know who they are, you know what they care about, you know what they said.

And it's all through a simple chat interface. You don't have to, you don't have to like, go between tools, stitch data together. So it's all in one place.

Jay Myers: Have you heard of the, you probably have, of course, the thousand true fan

concept.

Andrei Negrau: Yeah. Like when you start a company and like all you need is to reach.

Jay Myers: There, there's that. It's a little bit different. So the con yes, there's the get first, get your 1,002 fans. But there's this concept, I forget the, if you search it on YouTube, it'll come up with the guy that came up with the principle. But the, what a lot of brands do, this is what comes to my mind when I think of Sienna.

Like a lot of brands, they try to solve for everyone. It's like people have

different complaints and they try to solve for all I don't know, say you sell a dog bed or whatever. And some people complain about one thing, some people complain about another and you, you try to solve for anyone. But anyways, his philosophy was that find your true fans.

And so you go to the super users, like the people that are leaving five star reviews, the people that are going to bat for you on social media. If someone says something bad about your brand, they're the ones backing it up. They're the ones posting pictures. They're the ones. Referring friends to buy like you find your best customers and then you interview them.

So Sienna could maybe help with this, but his approach is maybe you only have 17 of these customers. Maybe you have five, maybe you have 50, but everyone has some super fans. Then you interview them and you learn. Everything you can about that customer, you learn why they bought it, exactly how they're using it. 'cause why do they love it so much? Why have they referred 10 people? Why are they raving it about on social? Why have they left five star reviews? What are they, how are they using it differently? What settings are they using the product in? Like, why are they such big fans? Understand them and then. Go out and find more people like them versus trying to fix it for everyone.

And you end up kind of with this bland product, and then you get this, it's a little bit more of a narrow focused product, but it solves a very real problem for a certain segment. And then you, those fans become your growth engine, right? So my mind goes to. Maybe there's learnings you could get about if you could find your true fans.

So find all customers that have ordered more than a certain amount of times, have left positive reviews, have done certain actions. Tell me everything you can about these customers. 'cause I want my marketing to get more of those types, you know? Oh, what's your thoughts on that? To me that seems like that would be something I'd love to try.

Would I assume that would be possible?

Andrei Negrau: Yeah, I mean, you're my, my wheels are spinning now 'cause I'm just thinking, I'm literally thinking, wow, like how much of this is already doable with.

Jay Myers: Could call it the true fan module because that's what you wanna build. On top of you always wanna, I'll put the link in the show notes to that video. It's about a 10 minute video. I wish I could remember the person's name who came up with the concept right now, but I'll make sure it's in the show notes.

But they look at all these businesses and the healthiest businesses are always built on top of their true fans, not trying to fix everyone's problem, and it becomes a mediocre product versus a cat bed for people who have a certain type of. Cat or whatever, and it's like the best in the world at that, or whatever the product is, right?

What does it look like? So something that goes through my mind as I'm hearing all this is it feels to me like even if I have people doing customer support like humans, I might still wanna use Sienna because of the insights I'm getting from it. Like, how are most brands using it? Is it a blend of human slash ai? Do you have any that are doing, like humans are still doing the support, but Siena is this data layer that they can learn from what? It feels like you're not trying to say switch your whole support to ai. It doesn't feel like that. Like how are brands, what's that mix look like?

Andrei Negrau: Yeah, so we're seeing a wide range of models when it comes to how brands use Siena and all of our different agents and all of our different products. There's biggest, by far, the biggest chunk of our customers are using both the support experience is part by Siena and now the intelligence. ISS powered by Siena.

And that's because somewhere in their leadership, somewhere culturally, they made this decision that, hey, we want to be as lean as possible, and we want to be as smart as possible with what we do. That allows kind of, it, it's almost like you building the foundation or you're building, in this case a plane.

You want everyone to be on board you know, fly the plane and go really fast. It's not just Hey, we're gonna build one. One room and then the rest is kind of like outside. So they love the idea that they can do both. And one operating principle that we share inside, like in, in Siena, and I always say this one plus one equals three.

So if you're really serious about running on ai, like becoming an A native brand, it's of course easier and best to use both. But there's also, we also work with brands that have very complex. Systems in the backend where deploying these agents take a little bit of time so it can take months until you actually reach a level of Siena really being productive in your support experience.

Jay Myers: Right.

Andrei Negrau: So with those, what we see is they already start leveraging the intelligence piece from day one. They you know, they take their time to launch the support experience because again, you have to connect different tools. Some tools are available out of the box, like your, you know, your typical stack, like your subscription providers, and then your, obviously Shopify is available, but a lot of enterprise brands, they have very bespoke setups.

So with those, they would kind of like stagger. So first do intelligence because it's available day one and then. Deployment will happen over the next few months. And we do have we do have a handful of folks that are using just CNA intelligence for the time being and they see value because your point is this intelligence layer that sits on top of what you already have.

So we, we see a wide range, but the pie chart would most of this lean towards use both the support agent experience and UCN intelligence?

Jay Myers: Gotcha. Okay, so you mentioned when the, when you are setting up. You said something like, as you get more comfortable with Ci Nena, so is that the typical on you, you have Siena answering 5% of tickets, then 10, then 15, is that how it.

Andrei Negrau: It's a great question. One of the things that is unique about, I think support AI in general is. It really mimics the philosophy and the processes and the the tool stacks of each individual brand. So sometimes I'm being asked, how long does it take to to launch cno? It's really hard for me to give like a, like of course we can look at the average, but the average may not apply to you.

So we see, we've seen brands go live and have a productive agent that already resolves more than half of the support conversations in two weeks. And it's doing it better and more consistent in terms of quality than human agents. That's another thing. That's that's another thing that's available and possible with Cena intelligence through spin up a QA agent.

So not only have your support agent, you can spin up and use Siena QA as a way to qa not just Siena, but your whole team. And then you can see reports, how is Siena doing against humans and how are humans doing amongst themselves? So we're replacing with. Completely replacing these pre AI QA tools because everything is integrated, so that is very important when it comes to ai.

The fun thing that it's like an interesting perspective is there's two types of agents in general. There's human in the loop and there's fully autonomous agents. Human in the loop cloud code is a perfect example, or. It's an agent, it's ANM that does some work, but at some point comes back to you and asks you, okay, do you wanna continue?

You wanna, so you're always constantly prompting that agent. The other type is fully autonomous, aI agent for support. If it's running on larger model it's in practice, fully autonomous. What that means is once you configure it in a specific way, once you train it, you give it tools, you, you give it access to what it needs, it will carry on the task fully autonomously with no human intervention.

And for that, it's really important that. When we deploy the agent, it really works well. So we have extensive testing happen. We have we have built a lot of ais and systems internally to make this deployment as soon as, as fast as possible and as good as possible. But accuracy and making sure that the AI is doing the best work possible is incredibly important.

And that's why. You know, in the beginning phases of ai, there's a lot of companies trying to do AI for support. This is back in like 2020 3, 24. There's so many different players trying to do this, and if you look around, there's not that many that are still around because it's actually quite complex.

So deployment can be a matter of days or weeks or it can be months, it can be truly months. And then there's another variable, which is how big is your, like your operations in the backend already. Like how many different tools, how many different agents do you have? How many systems, how many brands you have.

That's another thing, like if you have multiple brands, so all these, it's almost like different complexity knobs. We now, you know, when we, someone comes to us and we start a process, we start a, we started a deployment process together. We already. A pretty good understanding complexity. We can estimate, okay, this is probably gonna take a month, two months.

Internally, we're quite quite aggressive around. Building more and more tools, more and more ways to make this process more seamless, more easy. But it also requires a lot of human judgment on the brand side. That's why one of the things that I always share as an advice is try to set up your goals.

Try to set up your structures so that it aligns incentives with deploying an AI and having an AI run in your company. If this if something like Siena, it's just like a cute pet project for someone and doesn't really have. Any real tangible goals or anything like that. We've seen these deployments fail in the past because no one really cared.

Is it? Oh, yeah. Like we're, we use ai, we have a chat bot. It's okay. That is a very different mindset versus no, we need to reach this percent of automation rate in the next 90 days. Who's responsible? How do we, you know, how do we work as a team to get there and then everything works better? Once you have that sort of mindset.

Jay Myers: Speaking of which I saw, I think this was a case study somewhere hex cla increased automation 65%. I don't know if this is like outdated and maybe it's even higher now. But they, and they cut ticket. It's to humans by 20% in 60 days. So they seem like a obviously Hex Cloud is an amazing brand. They do a lot of things right. What did they do that was so great to get these results so fast that maybe other brands skip?

Andrei Negrau: It all starts with the operating team. Who's in the driver? Who's in the driver? Who is in the driver's seat for getting Siena up and running?

Specifically with Hex Cloud team, they have an incredible team and we work really close with with a few folks there that, and there's obviously an owner that oversees Siena, but their whole team is bought into ai.

And that made it, it's almost like the like a domino, if the team and everyone is bought into CI is bought into ai, everything gone. Runs so much smoother versus if there's friction. If someone out there doesn't really believe in AI or feels like AI is just a crut or it's just like a whatever trend or it's not gonna work, you're never gonna see incredible results.

You are never gonna see full, you know, full deployments happen. So in, in a case of HX Cloud, they have really good team, incredible product, and they really care about both customer experience and ai. Those are, of course, the results of various different. Work streams and various different things that lead to that.

Plus they're very good with with sharing feedback. Back when, a few months back, we CN our video processing. So we're able to process videos, not only text, not only images. CNN can actually, you can send it, upload the video, and Siena will know what's in the video.

Jay Myers: I saw you post that recently on a LinkedIn, or you or Lisa posted it so I understand. So a customer, if they're having a challenge, they can record it, either a screenshot of something on the website, or can they record a video of the product like,

it's broken here, this doesn't work. And then send it,

and Sienna can understand what's happening in it.

Andrei Negrau: yes.

Jay Myers: Amazing.

Andrei Negrau: So you can think about replacement. You know, let's say you, you know, let's say, you know, you got this glass, you ordered some glasses, and it's broken here. Instead of just sending a picture to say, guys, hey, you know, I got this. It's like all over the place. Send that and then see out. We'll, we.

Just like a human, interpret that video and make a judgment call. And that's another reason why Hick Lab was able to increase their automation rate. They, of course, the pens are highly, you know, they're very tactile. You have to, you know, you have to show Hey, there's this little bit of a scratch here, or whatever.

You know, something's happened here. So video has helped them and us quite a bit with with just making sound even more

Jay Myers: Yeah. And then I guess the million dollar question, can it detect if it's an AI created video? Because apparently everyone's just making images and videos of their products broken to get a refund. These days, I don't. I imagine they can. Is there ways to detect.

Andrei Negrau: The best way to think about image and video processing is if a human could the AI could as well. Sometimes maybe even better. Sometimes maybe even better. There's no, I mean, not today, there's not like a built in watermarking system or something like that, but who knows? In the future, I would say, you know, high high likelihood that the AI will get better at understanding if the video is generat With ai, we do see CCNR working really well with fraud cases.

Like maybe in this case it's not a video, but it's an image, you know, someone generates like a fake image of a product that's broken, right. I don't wanna train the ais, but you could quite easily do that. We've seen customers use it and they don't have, again, they don't have problems.

Like you can absolutely, like you can

Jay Myers: Does Siena have any insight? If a customer isn't a bad actor on one store and then on your store, do they have any intelligence in that or do you keep a fine line?

Andrei Negrau: So that's a great question. Another intelligence layer with a new product that we're building that will have, if you use that product, you'll automatically enroll your customers in this cross, cross brand. Of course, anonymized, but it will be almost like, like a watch, like a watchdog. So you'll be able to see, hey.

Jay Myers: has submitted 17 returns with over here, yes.

Andrei Negrau: Yes, and then we actually see a lot of input signals from brands. Trying to find ways to combat fraud by either telling Siena, Hey, look at if they, yeah, look, if they initiated multiple returns, look, if they used in the past multiple addresses, right? Because what they do, they would always change their addresses or like use, you know, shore street or N dot Shore Street.

So there's all these things that people got really creative with. I think the good news here is. You'll be able to, you'll be able to do a lot more with AI that you can with humans, just because AI has infinite attention. You know, it just works in, in, in kinda like it can work in parallel with millions of customers if it happens.

So yeah. We're looking at ways to bring these sort of inputs like fraud detection and stuff like that. It will be more and more important. Yeah.

Jay Myers: What are you most excited about for Sienna for the next 12 months? If I have Lisa on again in 18 months or so, looking back, like what do you of what you can say, what are you excited about for where Sienna's going and what the future looks like?

Andrei Negrau: You know, there's this point in time where we realized that we can really, our constraints, our actual constraints as a company have almost disappeared overnight. This wasn't the case a year ago. We had many constraints. We still have constraints. We specifically don't wanna raise a ton of money.

Or we don't wanna go the path of just keep raising for the sake of raising. So we have constraints, but in terms of what we can build and where we can go and where we can take our customers, it almost vanished. And then what really what, there's a few really interesting projects that keep me up at night that get me really excited.

One of them being the whole idea of building your brain inside Sienna, like building your company brain. Inside Sienna, we're about to launch depending. Podcast goes out. We're building Sienna docs. And what Siena docs really is your brain inside Sienna, not only your brain, but your, so everything internal, your SOPs, your documentation.

So think about replacing your Google Docs or your notions or gurus of the world. And on top of it, you're also building a help center inside Siena. So it's a single place to have all your data and why this is exciting. It's because for the first time an AI agent like Siena or, you know, multiple agents in this case don't exist in an external kind of in an external world where, hey, you just spoonfeed like a little bit of data here a little bit.

But in fact it will sit on, on, on the data that your team runs on. And we're planning integration with GitHub. So we really lean a hundred percent into how do we make Siena deferral the first true operating system for CX that's connected to your cloud code. It's connected to your data and everything is in one place that is.

That is incredibly exciting. And if you're familiar with skills like what we're building is a repository of skills. So you can now use. All the use cases that we talked about in this podcast, you can basically package them, put 'em in a skill, put 'em in a playbook and have your whole team run on it.

That's pretty exciting. It opens up the world for so many new things. Even, you know, things like hey we're launching a new market. Hey Sienna, we're launching a new market. Can you check out our current policy? What's your current returns policy? Hey, so check your website.

Is gonna check your internal docs.

It's gonna check the actual agent guidance. So it's gonna say, okay, this is where we're at. Can you now create a policy for the UK or for Europe? So everything is in one place. You don't have five systems that are not talked, not talking to each other. They just live in a complete silo.

Everything is now unified by CNA in the backend and in the front end. And that's really exciting because it, it means you can get even more insights, you can do even more with ai. And it opens up. A platform essentially to build on top of Siena. So that's quite exciting. Among all the things that we're building

Jay Myers: Amazing. And I imagine as you expand to different. Internationally, you Siena can interact in different languages and you can maintain those, your docs and SOPs in one language and it can then

interpret however. Yeah. Yeah.

Amazing. Just before we end here, is there like brands listening, wanna look into Sienna is it for, do you need to be a certain size of store?

Do you recommend it at once a store is at a certain level is there a path for anyone to get started? Or what, where would you recommend stores start with Sienna and how do they get started?

Andrei Negrau: So how to get started Siennas, ie. NA cx. So single end CNA cx book a demo, get in touch with our team. We pride ourselves in just trying to learn as much as we can about you and see if it's a good fit, if we can help you. Where CNA starts to make sense for a brand, we have different agents. So first of all, we have different agents that do various different things and more agents are gonna come online.

But generally, if you look at your customer conversation data, so not just support tickets, but this includes social media, this includes reviews. If you're already processing over. A month probably. It's a good time to think about it as and it also depends on how fast you're growing. If you're growing incredibly fast and a projector that you're gonna 10 x or 10 x reach out as soon as possible because that's the success story that we've seen with with this brand.

They reach out when they were growing. They're not as big as they're today, but they're growing. So it was just the right time as they were on their growth trajectory. So yeah, if you're growing fast, yeah.

Jay Myers: probably something to be said for if you are getting a million tickets a month, it's a lot harder at a certain, at a, on a growth trajectory it might be. Better to get ahead of it early and then scale with Sienna versus bringing it in later. I would imagine that's an accurate statement, right?

Andrei Negrau: It can.

Jay Myers: it not?

Andrei Negrau: It can be, but not always. Not always. Because with the tools that we have in place right now, even if you're doing a million tickets a month or a million tickets a year, we can see what those tickets are about and then we can kind of like practically, we can be one step ahead. We don't have to, back in the day, you have to manually swift through tickets.

Even for us, when we launch it out, we have to manually like. Check what people are saying just because there was not enough fidelity of data. But now we can just ask Sienna Hey, run an analysis over the last 30 days. What what are the main topics? And then, you know, you start training the agent as more like you're working backwards for the most, from the reasons that people reach out the most.

So yeah tools are here. We're excited to work with brands that have strong opinions about how they think CX should look for them. But also we're here to share what we see other brands do. We work with brands that are doing billions and dollars, and we work with brands that are doing $20 million a year.

Jay Myers: Mm-hmm.

Andrei Negrau: We have this knowledge, and we have this experience of seeing go across the board what works, what doesn't. And the one, one constant that doesn't change is the faster you start, the better. And it's a mindset like it is totally a mindset. Also, it's important to start.

But truly try to go all in versus just try an AI and you know, it didn't work, but we're here and we're excited to learn more and share about automating your cx, but also building your intelligence layer.

Jay Myers: Yeah. Amazing. I think. In my opinion, in five years, maybe sooner 'cause things move fast, maybe a few years. This will just, this will be the norm. I think every brand will be augmenting CX with AI or fully using AI in some way. So I just think for brands listening, get ahead of it now. Like now's an opportunity to move faster than your competitors.

I think when new technologies come out, it's always great to jump on them because it's a competitive advantage. Now that might be. Table stakes in five years, but now it's a competitive advantage. So Ena is, I think, probably the best platform I know to do this. So definitely check. Check them out. Ena, with one N.

'cause I have accidentally typed in two Ns many times. Is that domain for sale? Can you buy it? Or is someone, they won't sell it.

Andrei Negrau: We have a host of domains out of which I think, I'm not sure if with nx I have to, I actually have take a

Jay Myers: Okay. But we'll make sure it's CNL with one x. Check him out. Andre, thank you so much for your time this was a ton of fun.

Andrei Negrau: Thanks, Jay. Excited.