Loading summary
Transcript71 lines
- [00:00]
A
I can take that list of names, give it to an agent and say, build me a database with a photo from Wikipedia, a brief bio and a summary and 10 minutes later I have the database built.
- [00:09]
B
You look up the company Snowflake. It's huge. Multi billion dollar businesses make a lot of money. Data warehouse is huge because when you have all of your data clean in one place, you can make better decisions. Once the agent is there, you just have to change the way you work, right? This is just about resistance to change, to just get yourself over this change resistance. And if you're not lazy, it's a game changer.
- [00:29]
C
You get up in the morning, you open your slack, you just say, add the rates. How were my sales yesterday? You get an answer. If you see that the sales number in a particular country are below what you think should be, you can just ask a follow up question. Sales seem low. What happened?
- [00:43]
A
Someone's going to need the central repository. And it sounds like that could be the serious accusation. All of these diversions have to talk and bring data into you and then someone has to synthesize it, clean it, and then give the Ridge version to a Ridge executive to make a decision. And it's just a very exciting future.
- [00:58]
B
If you're not making decisions on data, you're just using your gut. And honestly, that's just overconfidence, period. You're many people that are just like literally flipping a coin.
- [01:06]
C
That's one of the things that I've been doing internally for the last 18 months, doing these AI audits of workflows and processes internally, documenting every step of the process and then looking at each step of the process to see whether that step is AI replaceable or that step is AI assisted. Get your AI foundation ready today. Agents are coming. You will see a serious unlock in productivity. Don't let your team struggle with spreadsheets. Get them on a foundation. Give them their time back. Unleash their creativity to help grow your business.
- [01:37]
D
Welcome to the Operators podcast. My name is Mike Beckham and we are proudly brought to you by Fulfill After Sell, Rich Panel, North Beam, Sara's analytics and Postscript. We are a community for entrepreneurs that are building things and if you want to be a part of this community, you can listen to the podcast. But you can also go sign up for our newsletter. A ton of awesome information in there. We also partner with E Commerce Fuel, a forum for you to connect with other entrepreneurs that are building businesses where we can learn from one another. So without further ado, onto the pod.
- [02:15]
B
Dealing with big box retailers means EDI connections, and that's often a trigger for needing an ERP system. We've been using EDI connections to Costco forever, and the only way that we've really solved that problem is to make it seamless is through Fulfill. EDI adds complexity to everything you do, and Fulfill solves that complexity with their connections to their systems. You need Fulfill to move from being just a D2C brand to being a true multichannel brand, because big box retailers are going to require you to connect to their systems using edi. Let me tell you, it's way easier if you do it with Fulfill.
- [02:48]
A
Welcome to the Operators podcast. Krishna, we could do this in person. All of us are in LA right now. How's how you join California?
- [02:56]
C
I'm doing very well, Sean. Last. Last couple of times we did this, it was 10pm for me, and right now it's 7am for me.
- [03:02]
B
Jet lag.
- [03:03]
C
But yes, getting looking forward to getting started with the conversation. Excited. Would have been loved to, would, would have loved to have this in person.
- [03:11]
A
But nevertheless, dude, you know when, when the operators podcast calls, if it's 10pm or 7 in the morning, we get it done, guys. Jason, how are you feeling?
- [03:20]
B
I'm very happy to be here. I'm not feeling great because I've had a cold for like the last week, so I'm gonna be going on mute while I hack up along. But Krishna was in my office on Monday. He had his team in. We had a bunch of meetings. We're cracking the whip on saris. They're doing some pretty cool work for us. So I'm excited to have Krishna back on the pod.
- [03:40]
A
Okay. I love it. Today's episode is all about data, data integrity, the importance of data. The reason why we're doing this episode is because with AI rolling out everywhere, that's one of the initiatives that bridged this year's AI everywhere. We're finally getting validation that, like, all the data we've captured is important. So we have Krishna here to talk about what to do if you're a $50 million brand or below, what data is actually important, what you need to be focused on, and then what do you unlock with AI when you get all this data ready? So, you know, if you're a brand right now and you have no data integrity set up, if you're, if you're just not tracking stuff, you're not even downloading CSVs from Shopify, this episode's for you. We're going to help you unlock things you can do and really try to demystify all the AI stuff. I think there's two camps of people right now. You either hand wave that AI is going to do everything or you hand wave that AI totally sucks. Obviously there's something in between there, right? It's not doing every job in my business right now, but I am making ads and I can show you guys some amazing AI creative if you guys want. I've been cranking those out myself. So we'll do something in between to figure out what's really going on in the world of AI right now. Sound like a good episode.
- [04:52]
B
I love it. And, and just Sean, just real quick like you know, you in your organization have been, you've been super focused for months now, you know, pushing people to get into it and to really leverage AI and, and I think people are doing it. I'm actually pretty impressed. I'm starting to feel better about it. Especially with the new Claude release. People are actually embracing it and are trying to do cool stuff with it. I think, you know, for me going forward, when I'm every hire I look at now, it's like, are they going to embrace AI or not? Like I'm, I'm. We just hired a new Middle east country region manager because we're, we're selling in Dubai now and I, before we, we made the offer, I said I need you to send me two things and you send me a five year plan and I need you to because you know, people are doing like these two year plans. I'm like, it's not that exciting. Let's see what it looks like over five years. And then I said, explain to me your feelings on leveraging AI for the business. And you know, he basically wrote like every, every. Right. Wrote sort of a position paper on everywhere within the E commerce stack. Like we should be using AI and just taking the, just taking, making the effort to do that I thought was really nice. And so it's got to be like table stakes for everyone at this point.
- [06:18]
C
Yeah. Talking about table stakes, Jason. So I've been on this AI journey personally for the last 18 months. So as a technology company, we have to transform internally. I have a 200 people team that I have to get them to adopt AI Think AI first. The culture in the company has to change because the customer expectations are changing. Right. You expect things to get done faster, better with the same set of people. So there is the internal aspect of it which is how do we change culturally as a company? Something that I've been experimenting with for the last 18 months and externally to customers. Customers are going to expect AI output from us. Right. So, so for us it's on both sides. One is on the people side and the internal processes and how we can do more with less by becoming AI first as a company internally. And the second part is what do we put in front of our customers so that they can become AI driven as well. Right. So having gone through this journey, it's been interesting, right, because not everybody is adopting AI at the same pace. There's some early adopters, there's some non believers still, which is very, very surprising. So having gone through this journey and getting the teams to transform, there are some learnings that I would like to share if time permits. But yeah, happy and looking forward to chat about what we could be doing here with AI and how exciting times are right now.
- [07:49]
A
Yeah, the non believers, let's start there, right. I think most people listening to this probably have used AI. I mean everyone definitely has, right? Even if you don't want is going to be Google knowledge box summaries at the top of every search. I can't believe someone being a non believer in 2026. Like if you used ChatGPT or Dolly four years ago, I totally like get be like oh this is really cool, right? And then moving on with your life. It is totally different now. It's a different product. Calling it AI is actually like unfair. Putting it all in the same bucket. I'm using agents now and like agentic work is awesome, right? And I can give you guys an example of what that is. I got a list of famous people from a talent agency. These are people we want to book a commercial with, right? And it's, let's call it a hundred people. I don't know any of them. By reading their names, I can't picture their face. Right. So I could individually google every person and look at them. Or I could take that list of names, give it to an agent and say build me a database with a photo from Wikipedia, a brief bio and a summary and 10 minutes later I have the database built. It could be a thousand photos, it could be 10,000 photos. Right? Like it is, it is doing the work of, you know, junior level college employees right now.
- [09:05]
C
Yeah. When I say non believers, Sean, I'm talking about folks who are still not convinced that they are going to be talking to agents first before they talk to humans next. If I look at the roadmap for service analytics over the next two years, I can assure you, or even the next one year, I can assure you that most of the first conversations or interactions that customers are going to have with, with my company are going to be with the agents that we built first. And then if the agents are not delivering the right answer, then talk to, then pull in a, pull in a human. Right? So, so from, from my standpoint I'm looking at non believer as somebody who is still not convinced that they are going to be agents are going to be front ending them and they'll be interacting with agents. So it's not necessarily all whether they're using AI or not if that makes sense.
- [09:59]
B
They're not convinced yet. Right. And they're either going to be convinced and if they're not, it's because of either two either they're, they're lazy or scared.
- [10:08]
C
Right?
- [10:08]
B
They're, they're lazy because they just don't want to take the initial, initial time and effort to, to figure out how to use it. And it's, it's actually so easy, right? Like once the agent is there, you just have to change, you just have to change the way you work, right? This is just about resistance to change and you literally just need to spend I think like like 10 minutes, 30 minutes to just get yourself over this change resistance. And then, and, and, and if you're not lazy, I mean it's, it's a game changer we were looking at. I mean this is always what I've wanted. I've always, I've been saying this since I was a banker. I used to use huge databases as a banker, FactSet Capital IQ to get information about companies, right? And we would have to build these massive models. I had these people building massive models, massive PowerPoints and all I wanted to do was be able to just write a question to this database and spit it out. Right? And that's exactly what's finally happened. And this is what I've been talking about with Krishna like from the beginning with Saris. What I want back when I, back when it was, when it was at, @factset in FinTech, back when I was a banker, it's like what's a solution for like C level executive to like really easily get answers. And that's what this does for me. Like it's, it's, it's just like absolutely insanely good at this point. I'm, I'm, I'm so excited about where, where all of this is going. Krishna,
- [11:37]
D
if you're scaling an E commerce brand today, ads alone aren't enough. After Sell focuses on the one moment that every brand already owns after checkout and turns the post purchase moment into more profit. Monetize every order with post purchase offers and thank you page experiences without disrupting checkout or hurting conversion. Enterprise grade tech used by Gap, Ticketmaster, Macy's and Target. Now driving results for brands like True Classic, hexclad, Ridge and Jones Road. I would know. This is the reason I ended up buying three pans from Hexclad instead of two. After sell has already generated over 1 billion in additional revenue for e commerce brands. Revenue that doesn't require more traffic or higher cac. So check out after sell and tell them that the operator sent you.
- [12:23]
C
Absolutely Jason. So I'll give you some real life examples for us. Maybe I can talk about stuff that we're doing internally and then we can talk about things that we're doing with customers externally internally. For example, I have a large chunk of my team sitting in India and very often customers ping us at 2am India time, 3am India time asking a clarification question saying hey, my numbers in this dashboard seem a little bit off. Or how do you calculate this metric? Right now to answer that question I need to have a customer success engineer who is trained up on our technology can take that question, understand how to find an answer to that question, right? Or consult the right people in house and then get back to an get back to the customer with an answer, right? So the average response time so we we built our own internal data lake. So whatever I go talk about building a single source of truth for brands applies equally to every single business, no matter the scale of the business, right? So we have our own internal data lake where we have all the interaction data, everything that we are loading up into the system. What we have done is we have measured what is the average response time across all the questions that we are getting from customers and the average response time of questions that we get during our work hours, which typically are until midnight, is less than 30 minutes or 30 minutes to one hour. But if the questions come to us after 2am, the response time increases to eight hours, right? So we are going from being prompt at 30 minutes to one hour in terms of response time to being not present when the customer is perhaps sitting in a board meeting or in an executive meeting trying to review numbers and they have a question that comes up and here we are not able to respond to them on time where we are going. And we are testing this in beta. Internally we have some agents that we have built, we call it IQ Engineer and IQ Analyst. So you can ask the same Question our agents respond to you in real time or in near real time in less than 60 second. Response latency on these questions around. Okay, how did you calculate demand revenue, for instance? Right. It doesn't require a human. Agents can answer that question and if the conversations go beyond a certain point, then a human can get engaged. Right. So that's just an internal example. An external example that I can think of is I recently met a product team and the product team is trying to understand for the product that they're designing or the product that they've come up with, what are the conversions on the site, how many, how much traffic am I getting to these web pages, how much is my marketing team spending on campaigns to drive traffic to these products, how much revenue are we getting, what discounts are we giving, et cetera. And they just don't have answers to those questions. All of that would become a single text prompt with answers coming back to these users in 60 seconds. Right. Imagine what that can do to your business where any stakeholder in the company can ask a question, get the answers of their data without waiting on their data analyst, their data team, et cetera, but instantaneously so that they can take a decision and move forward.
- [15:46]
A
Yeah, you know, let's go, let's talk about the internal example first. I think we can all agree and all the listeners can agree in 10 years you will not hire customer service reps. Like it is so obvious that email based customer service. Where is my order for E commerce orders? That will not be a human job in 10 years. And then it's just in two years
- [16:10]
C
or maybe even less than that actually.
- [16:12]
A
Yeah, yeah, yeah, but I'm warming everybody up. Right? So I'm saying it, we all agree 10 years and then it's just how aggressive are you? Is it five years? I would say yes. Is it four years? Is it three years? Is it by the end of the year? You're never hiring another CX rep. Right. It's like the technology is moving so fast, you can just so obviously see that work being done. It's the same thing with self driving cars. I bring this up all the time. Time. It's like you're not going to drive your car in 10 years. We can all agree that it's like the cars can already drive themselves. And it's just like where on that curve do you think the technology actually happens? And with any repetitive knowledge based work like customer service, I think, I think it's literally by the end of 2026, you're not hiring new People for that role. Now, the external thing, let's talk about that first. We should, we should. I don't think we've ever done a good explainer. What is Saris Analytics? So it's a data warehouse. Okay. And Saris analytics as a company comes in and they clean and set up your data warehouse. Okay? So you have data in Shopify, you have data in wholesale, you have data in fulfill your erp, you have data in Amazon. The problem is all that data is not clean. What I mean by clean is my title for Amazon for a gunmetal wallet is like best wallet ever. You know, it's an Amazon title, right? So if I, if I just try to figure out how many gunmetal wallets I sold, I would have to go in there and have to manually think about that, right? Sales going to come in and they're going to map those things and clean it up for you. And then they're going to give you a box that is all of your data updated every hour in real time that is clean and custom designed for your business. Christian, is that right? Is that what TS analytics does?
- [17:51]
C
Yep, that's what we have been doing for the last 10 years. We are taking a couple of steps further than that, Sean. We are looking at Cyrus as more of an AI foundation and an AI workflow engine for brands and agencies. And a data warehouse becomes a foundational element here, right?
- [18:10]
A
Yeah. Let's not talk about the future yet. I just want to get everyone familiar with why they would need a data warehouse and then we could talk about what we could build on top of it, right? So like at the very base level, if you're a brand, having a data warehouse is important because as far as soon as you start selling on more channels, as soon as you have a couple of years of historical data, somebody changed a price way back. Someone changed a sku way back. You have to get all that information in one spot. And right now it's, you know, ancestral knowledge. Like you just know because you did it right. But when you're gone or you hire somebody else, it's hard to translate that information. The data warehouse makes that easy. The advantage of having a data warehouse in the AI age is now you have all your data in one spot, you have your marketing data in there, you have your spend data in there. And now you can start building agents on top of it, right? Or reports on top of it, or query into it, right? And now, Krishna, then you tell us about you have the data warehouse. Where is the future of Saris analytics going.
- [19:07]
C
The future of service analytics is agentic, Sean. So I imagine a user coming to Slack and asking a question saying, okay, I just launched this product last week. What's my performance? What's the performance of the product? Right? And I expect them to get a summary of the product performance since launch on vital parameters that can be used to assess whether the launch went well, did not go well, or doing average all in a matter of seconds. What does that mean from a practical standpoint?
- [19:38]
B
Right.
- [19:38]
C
If you think about it, when I launch a product, the orders are getting logged in Shopify and Amazon. So my sales data is in these two platforms at least when there is traffic, advertising campaigns or money is being spent. So that data is sitting in Facebook and Google and all these other places, somebody clicks on the ad, lands on your website, the conversions on the page, the activity on the page is logged in GF4 or Edge Mesh or one of these tools. So for a product owner or for a marketer or even a CEO for that matter, to understand how did my product launch go, you need data from all of these systems to understand what is my spend, what is that spend leading to from a conversion standpoint and what does that mean to us from a revenue standpoint? Right. And how does that face against the targets that we have set for ourselves? Right? So getting that summary today is so hard. Like it takes multiple people going to multiple systems. All of that is going to get reduced to seconds actually going forward. And that's the direction that we are taking our customers.
- [20:44]
E
Every SaaS company says they are AI powered, but very few can explain what it actually does for the revenue of my brand. This is why postscripts approach stood out to us. They don't just build AI for demos or buzzwords, they built it to drive real incremental revenue. Postscripts AI called Shopper. It shows up inside of SMS at moments and with real buyer intent. When shoppers are likely asking questions, hesitating, maybe even about to drop off. Shopper can answer product questions instantly, answer questions about fit, availability, recommendations, order issues, the kinds of stuff that people usually bounce for. This means more conversions, higher a of less loss demand. So you are driving more revenue and doing it more efficiently. Check out Shopper from postscript. We use it at Pela, which is why I am telling you to check it out.
- [21:29]
B
You need a data foundation underneath your AI, right? Like that's fundamentally if you don't have a good data foundation under your AI, it's useless. And that's why like quantitative people have been less Interested in really adopt, adopting AI, right until, until Claude's most recent release. And, and you could like really power Excel with Claude and do really good stuff in Excel with Claude. But even then you know, there's so much time going into compiling and scrubbing data and it's a very manual human task. And if you, you know, if you are running on more than just Shopify, even if you're only running on Shopify, I, you still, you're probably still using north theme, right, or using another MTA tool and you're getting marketing metrics out of that. You have an ERP system. And so I've been a big believer in data warehouses for like 15 years. I mean if you look up the company Snowflake, right, like these are, there's huge, multi, multi billion dollar businesses that make a lot of money because this is like in the financial markets data warehouses are huge. And basically all throughout like large scale businesses, data warehouses are huge because when you have all of your data clean in one place, you can make better decisions. It's all about like making good decisions. So, so for us we, we pull in Shopify, Amazon, Costco, North Beam and many other sources and it's all in a great clean state within the data warehouse. And then we pull it out into dashboards, we pull it out into daily email summaries, data summaries. And now with AI you can get the, the real, the real power, the real leverage is actually like querying the database through AI to just answer questions that you have. Like you just, you know, I, I'm constantly going to, to chat and asking it questions now, but now I'm just gonna, I'm gonna be doing the same thing, asking questions about my data and it's just gonna allow me to make way better decisions. It's, it's like astonishing how, how good this is going to be.
- [23:44]
C
Yeah, a simple workflow there Sean. Jason, for you, like the way things are going to look very soon once we launch IQ for you is you get up in the morning, you open your slack, you just say at the rate for iq, how, how were my sales yesterday? You get an answer. If you see that the sales number in a particular country you know, are below what you think should be, you can just ask a follow up question saying the sales seem low, what happened? And the agent is actually going to get into your data, understand how the marketing, marketing performed previously, what could be some of the reasons why the numbers are lower than expected and it will come back to you with a better than usual, like better than expected answer. Sometimes you'd get surprised at the quality of answers these AI LLMs are able to produce. With the right set of context and training. You can actually ask a simple follow up question saying, okay, why is the sale slow? You'll come back to you with the summary so you can start your day interacting with an agent, getting to understand the current state in a matter of a couple of minutes and then get on with your day. You know what is the traditional workflow there? You wait for an analyst to put all of these reports together, usually manual, they're on leave one day, you're not going to get that report. And even if they are working on that day, you probably have to wait for them to compile everything and then ship that file to you. You get an Excel file and you have to open it, read it and comprehend what's in there. Imagine swapping that with a simple question and a nice summary that you can read in a couple of minutes and get up to speed. So that changes the decision velocity at a business. And having that power to any team that may not have budget to hire their own analysts is in my opinion going to unlock serious productivity for brands.
- [25:35]
A
Yeah, and I just, I just want to keep highlighting how people need to experience like agentic work to understand what's capable. Right. So right now you can use Manus. Manus got bought by Meta. If you spend money on Meta, they'll give you Manus credits for free. Ask for it and then just ask it to do something. It will go on the web. I mean I probably did this on the podcast. A year ago I had it ordered me Chinese food. But like now I can do like I told you guys, I, I had a task that would have taken someone an hour to do and I'm like, hey man, this. Just go in there and take photos and screenshots of all these celebrities and compile me and then make me a little web app for it. Like there is so much that can actually get done with, with the, the actual agents doing work right now. And I just want to make sure people experience that because it sounds like, oh, I'm just going to talk to the AI bot. Here's an examp Rich. We have serious analytics set up and we put all of our marketing data in there. Right. And it's very easy when things automatically port. Right. So Meta spend, Amazon spend, Google spend. Like there's APIs, you can pull the stuff in. But what do you do about podcast Spend. What do you do about TV Spend through Tatari. Right. So I give Tatari a budget and they spend the budget and then sometimes they don't spend all of it or they only spend so much of it. I have to know what they spend per day. So they email that to us. Our agent's just going to ask them that. Like our agent will bug them every day, hey, what do we spend yesterday on Atari? And then that agent will get that information and it will put it in Saracen Analytics. For me, that is a human removed from that loop. There doesn't need to be a human in that loop. It is getting data that is harder to find, but they can give it to an agent, can actually understand that. And then if Tatari was smart, they wouldn't have a human email my agent, they would build their own agent. So my agent's talking to their agent and just getting the spend data and you just start seeing how more and more results and actions in a company reduced down to, oh, an agent can probably do that. Right. So I just want to highlight agents, they're way different than normal LLMs or AI, and I'm getting used out of them right now. Krishna, why don't you go and show us Sarah's iq?
- [27:42]
C
Yeah. So let me give you a quick, quick sneak peek here. So here's an example of life at a typical brand. So if you are a founder, an executive, you're getting spreadsheets. This is usually what your team is struggling with. They might not be coming to you complaining about it, but they're definitely unhappy that they're doing some of this work. Right. So you have data sitting across all of these different systems. Advertising, marketing, marketplaces, analytics, 3pls, et cetera. You have your, either your leadership team or your analyst pulling data manually from these systems and then sending that to folks like Jason and Sean, who are then asking you questions or asking questions about data quality or asking you questions about how something was measured. And this is a tedious process, right? So someone goes on leave, you're stuck. What we've been doing all these years is we are giving these analysts and the leaders, operators in companies their time back by building what we call as the data foundation. Right? So what we are doing is we are pulling data from, we have connectors to 200 plus platforms. We pull data automatically in near real time from these different systems, load that data into a data warehouse, clean it, transform it, and certify it so that your data is ready for reporting, analysis or whatever work stream that you might have. Once you have the data foundation, what are the kind of questions that you can answer, right? You, you'd be able to understand your contribution margin by SKU, by campaign, by product, by product category, etc. So that will help you understand which are your top performing products, which are the products that are leaking you contribution margin, which campaigns are performing well, which ones are not, so that you can take decisions, right? You get some trackers that you can use so that you can understand pacing, projection and your actuals. You get a consolidated view of your inventory, you get customer cohorts so you can understand lifetime value metrics, et cetera. So once you have all of these, so this is your data foundation, right? This is your 101 right? Every brand, in my opinion, should, or every company for that matter, should invest in a data foundation where you have the single source of truth. Once you have the single source of truth, that's where the agentic layer starts to kick in. So we have a product called Service iq. And in Service iq, it's actually what is the goal, right? Any stakeholder, whether you are a data engineer or a data analyst, or a CEO or a cmo, we want you to get to the right answer from the right agent in the right language without you having to know how the data works underneath, all in a matter of seconds. How do we do it? So IQ actually comprises of six agents, each doing a different function, right? So there's an IQ analyst who is answering questions like what happened in my data? When you ask this question, IQ analyst agent is dipping into your AI foundation that we built for you and comes back to you with a response. IQ engineer is troubleshooting a data pipeline break, right? These APIs change sometimes due to latency. You don't get the data. IQ engineer kicks in when you say my data looks suspicious, can you go check what happened? IQ engineers, right? Scientist is predicting what might happen in the future. Data quality. Again, we want you to trust your data, right? Only if you trust your data is when you're going to use it, and only if you use it regularly is when your decisions will, you know, we'd be influenced by. So, so these are some of the agents that we're working on. They're all under the IQ agentic work workforce. And the way this works, if I were to give you a quick demo, so somebody who is signing up for IQ would get an interface like this where you can ask a simple question, right? So for example, give me sales for the last 60 days, broken down by category. Essentially what is happening behind the scenes is IQ is taking your query, connecting to the data foundation it's picking up the business logic that we have trained IQ on specifically for your business and your business context and generates the query and it generates a summary that you can quickly read and understand or get an answer to here. Right? So typically you'd be either trying to find an answer to this question on Excel, or you could just come here, ask and get a simple summary out like this.
- [32:35]
A
Okay, Krishna, but you know, the pushback I'm going to have for you is why are you building IQ from scratch? Why don't I just take the data and just put it into Claude 4.6 and have Claude do reusing models on top of it? What's the advantage of building the agent yourself?
- [32:55]
C
Yeah, that's a fantastic question, Sean. In fact, IQ is built on top of Claude, right? So we use CLAUDE internally along with a couple of LLMs. What Claude does. CLAUDE is a general purpose tool. IQ is a vertically specialized tool for E commerce. Right? IQ understands your business context and those are the agents that we are building. So we are building agents that understand your business context, what the metrics in your business are, how they are defined, what is the data saying about your business. So we build all of that context into iq. We then pass that context over to Claude and then let Claude do its magic. Right? So CLAUDE is generating the SQL for you. CLAUDE is helping us generate this visualization, CLAUDE is helping us generate the summary. But what IQ is doing is stitching all of those workflows together, passing the business context so that we get at least nine out of ten questions right when we answer a question, that's one. The second thing is, if you ever see any of the announcement that come out from, let's say, ChatGPT or Cloud, where they're launching a new model, the way the launch process works is they basically write a series of tests and they test the new model and the old model against those same set of tests, which are called benchmarks. And only after the benchmarks are passed is when the model gets promoted to production or general purpose use support by users in iq. Every time we get a question right, we actually ask the user to adjust it to thumbs up. The moment you hit a thumbs up, this question gets logged into our evaluation framework saying, okay, we got this question right? So the next time, let's say today There is Opus 4.6, Claude comes up with Opus 5, and we want to make sure that Claude is actually giving you the same answer. There is no guarantee that CLAUDE will give you the same answer. But by hitting the thumbs up here, we Create our own testing suite where we know what is the question that got asked, what is the right answer. And when we are trying to upgrade our models, we run, you know, parallel test against the old models, the new model, and only upgrade to the new models when we are getting the answers consistently right and beyond a certain threshold. You cannot do that if you are on cloud. Right. That would be a secondary. The third reason is user management and commissioning. You know, if everybody is using Claude on their own desktop, it's very difficult to build and maintain context across everybody. So a simple example of that could be, let's say I am downloading sales from Amazon, my colleague at Ridge is also downloading sales from Amazon and both of us are coming up with net revenue. My formulas are different, my spreadsheet numbers are going to be different. The same thing can happen in Claude as well. Right? Because people might be feeding in different contexts and they might be getting a different answer and the confusion increases. By going through iq, you avoid that confusion because now you have a standard definition of what electric means for a business or for a department, and that is governed by IQ quality agents, just to ensure that you get the right answer. So these are just a couple of examples that I can give you why IQ right out of the bat gives you a little bit more than what you could get with Claude. But the beauty of the AI foundation that we built for you is if you want to use Claude, you love it. You can absolutely connect Claude to the Data foundation and have fun.
- [36:34]
A
What's up, operators? Welcome to the Rich Panel ad read. Rich Panel has been a sponsor for over 12 months. I've been a paying customer for over 12 months and guess what, I just renewed the pay again for another year. We have cut our SaaS bill in half and Automation dropped our cost per ticket by 70%. Our CSAT has also improved from 88%, which is still really good, to 96%. Best in class. All powered by Rich Panel. I told them last year, hey, you guys need to do the same thing with returns. And now Rich Panel has a returns portal it's built to cut down your tickets and convert more refunds into exchanges. They do the heavy lifting, data import, self service, retention flows, team training, all of it. And it'll be live in two weeks. If you want to save 30% guaranteed on help desk and now returns, book a demo. Yeah, you know, it kind of sounds like you've built the skill that is, you know, E commerce analyst for Ridge into the agent. Right, but it's not really doing agentic work right now. It is, it is, it is. I guess it's going to your database and inquiring it, but it's not going to go out there and do work on the open web for you.
- [37:44]
C
It's not yet. So it is in our roadmap. So I'll give you a simple example. Let's say your campaign performance is dropping and we have one of the IQ agents detect that your campaign performance has dropped. What do you do next? Do you pause the campaign or do you reduce the spend threshold? There is some decision you might want to take right, from an IQ standpoint. By the end of this quarter we will have a few agents that will give you this signal saying, hey, this performance seems to be off, but we don't necessarily have to set up today to go take an action on your behalf. But that would be an extension of IQ down the line.
- [38:21]
A
Yeah, I have this debate with Connor right, about where, where the world's going because you know, Meta bought Manus to turn it into a ad buying agent which will do agentic work. Right. And House is building an agent and I'm sure Northbeam is going to build an agent. And what does this mean is that all of the work and data they have will have someone full time. Someone. It's going to have an agent full time prompting you to do what it thinks it should do. Right. And right now those prompts are going to go to a person. It's going to go to Connor or Jimmy, my VP of marketing. It's going to go to one of these people and there's going to be a market to build the Ridge agent who will take in all of these inputs and then actually summarize them for you. Because Hal might be saying I should spend more money on YouTube and Manus for Meta is going to say I should spend my money on Meta and Northbeam is going to say, ah, you know, they're both kind of right. You should probably do this. Who's actually going to be processing and grokking all that information and then giving you the unbiased thing to do. Right? And you know, it's going to be the Ridge agent, it's going to be the hexcloud agent. Someone's going to have to build an agent to receive all these inputs and think about the inputs as APIs. Someone's going to need the central repository and it sounds like that could be the Saris IQ agent. All of these different agents have to talk and bring data into you and then someone has to synthesize it, clean it, and then give the Ridge version to a Ridge executive to make a decision. And it's just a very exciting future. What I just described might sound like nonsense to so many people on this podcast, but I'm telling you, it's where the world will be in three years.
- [39:53]
B
It is 100% the future. What you're talking about is 100% the future. Right. Like, what you need to do is figure out what agents should be doing for you. Right. And how to, how to, how to implement that over time, you know, until it turns out that that kind of workflow system has changed. But I mean, once it's sort of set up that way, just like, you know, you used to, you type in www.in a browser. You know, the Internet is done through a browser. AI is going to be done through an agent, and your workflows are going to happen through an agent. That's just like the way it started. And that's, that's just where it's going to go. And what we're, what we're should be all figuring out is like, what agents do we need in our, in our business? That's what's happening.
- [40:44]
C
Yeah, absolutely, Jason. That's one of the things that I've been doing internally, actually for the last 18 months, is doing these AI audits of workflows and processes internally, documenting every step of the process and then looking at each step of the process to see whether that step is AI replaceable fully or whether that step is AI assisted. Right. Or that is something only a human can do. What we have identified is that for many of our workflows, the simple example I gave you around customer success is the first step a human does is actually go and tell the customer who reached out saying, give me a few minutes, let me check and get back to you. Right. That is a step that requires a human involvement that could be fully replaced by AI. Right. So we are doing these AI opportunity audits constantly across teams, documenting these workflows and then going after these workflows which have high roi. If we introduce AI and AI agents and going about doing that, and we are seeing some incredible success. There's something I'm happy to share with any of your listeners if you're interested in talking to me about it.
- [41:55]
A
Okay, so let's, let's summarize Saris analytics so far. It is a data warehouse solution. So you guys are, it's going to come in here and it's going to clean your data. And that's been the business the past 10 years is it's going to take all your data from all your e commerce sources, it's going to make sure they all are accurate, they speak to each other, it's going to bring in your marketing information, it's going to bring in your cogs, it's going to give you like the truest sense of daily profit, daily contribution margin. And it's, it's all set up on, you know, great infrastructure like Tableau and everything else, right? It's, they're going to come in here, it's, it's something like $10,000 a month, maybe a little bit more for, for our wonderful listeners if you're too big and they're going to set all that stuff up for you. That's the SARAH Linux the past 10 years. Right now what they're building is Saris IQ, which will take all of that data and just make it easier to access. You'll just be able to type in, hey, summarize this for me, just like you do with claude. It's built on top of Claude, just like you do with ChatGPT. It makes an LLM out of your data in your database just for you, just for your team. So it's proprietary, it's safe, and you can get your information back quicker, easier, with maybe more detailed prompts. The future is the sarah's IQ agents coming out, right where they will start doing work on your behalf, answering questions on your behalf. And I'm predicting it's going to go fight the meta agent for you, it's going to go fight the house agent for you, and it will end up getting, you know, the cleanest truth across all these different data sources. Marketing is not a solved problem, and that's probably why it's such a big industry, is that, you know, it's a trillion dollars a year of marketing spend, give or take, right? Between fees and everything else, it's a big, big part of the economy and half it's wasted. It's like it's the only part of the economy that is totally inefficient in that way, right? Like if you talk about oil and gas, their loss rate is 0.01%. Talk about Meta ads. Half of them are just letting money on fire. And in the next five or ten years we will solve that inefficiency, we will remove that, but it's going to be through agents fighting each other and really eliminating all the waste, fraud and abuse. Jason, what's your response?
- [43:59]
B
I mean, you look, you hit the nail on the head Right. It's, that's exact. It's exactly right. And it's funny, you've seen tech stocks, like, getting hammered, right? SaaS company stocks getting hammered because the people don't understand, like, or are concerned about the impact and that, that AI has. What is it going to do to these SaaS businesses? And the right SaaS businesses, just like the right humans, are going to learn how to leverage this, right? Like, so you look at the disruption in the workforce that people are afraid of, the people that are good at their job and leverage this to be even better are the ones that are going to be the winners, right? And the companies that are really embracing, the SaaS, software companies or tech companies that are really embracing, like, how do we, how do we integrate and, and just make it a part of our DNA are going to be the winners. And I think that's the, the real you. So you have to make that decision. And then it's like, how do you go do it? Right? Like, that's the hard part. How do you go do it? You know, you've been talking about this for a while. Like, you know, can see it, you can see that I got a demo and I saw how I saw agents, you know, working. But that Sean's point about that agent looking out versus looking in is a really good one. It's just like the speed of the rate of change here is like, so geometric. It's, it's crazy. It's just, it's, it's, it's wild.
- [45:39]
E
Longtime sponsor North Beam is launching Incrementality later this quarter order. This means that you can now have the trifecta of marketing measurement all in one platform. That is multi touch, attribution, media mix modeling and incrementality holdouts all inside of North Beam. You can automate that lift testing end to end, unify results with your MTA and your. Mmm, this is a lot of letters. But if you know, you know, and you can start to cut what doesn't work and you can scale what works and you can do this all with confidence. This is why this is such an incredible ad to North Beam. North Beam's incrementality measures what results marketing is actually generating, not just what they're claiming credit for. As a CEO, that's like music to my ears. Sign up now and you can lock in 50% off unlimited tests for the year.
- [46:24]
A
Yeah, and an earlier point you had, Jason, was that if you're lazy to learn this, like, you know, it could be a detriment if you're Lazy. You should be at the cutting, you should be at the front line for agents. They do the work for you. Like I cannot stress that enough. Like I'm like, I'm bad at a lot of stuff. The agents are good at stuff. I tell them what to do and they do it. It's like, it's like every CEO's dream, right? And now everyone can be their own little CEO. So if you're lazy, I really hope you embrace agents. And if you want to taste, look, right now I just have this tweet. It is, it is kind of like geocities building agents. And what I mean by that is like it's the early Internet and it's not going to be very useful and it's going to be replaced very, very soon. There's going to be commercial grade agents coming out from the Sarace analytics, the world, from the metas of the world, whatever. But you should still do it and play with it because it's a good hobby to have, right? Just start using Manus right now. Like you should, you should get credits from Manus. You should jump in there and start seeing what is capable, what the work you can do. Do a fun little hobby project and it is just cool what it can get done.
- [47:24]
B
It's a great point. The hobby project. You know, I was, Mike was tweeting about like Mike told me, you know, I, I, I reached out to Mike about open cloth because he was like tweeting about it, he was all about it, right? And people are talking about you know, security issues etc and I just wanted to know from like, like what his thoughts were because he was really all about it and he's like, I treat it like a pet, you know, it's like, you know, it's like his little pet project. And like what, what you're saying, Sean is like, you know, having like, like it's a little, it's a little hobby, you know, like it's something that is gonna, it's like a self improvement. You, by doing this stuff you're, you're actually focused. Everyone's focused on their health, right? Well this is like, this is actually going to really help you. You know this is like a health thing that you need to do is like make this a hobby. Nine months ago I was like, I was really not concerned about AI at all. And then like we talked to Craig, he came in, Craig Fold came in and we just started talking and he just showed it to me like that this is crazy. Like I got, I have this totally wrong like I was totally wrong. This is changing my life. And you know, and, and it has, and you just have to put the time in. And I think it's the, the one thing that I'm getting out of this conversation, Sean, even though it was, it was apparent to me, but not as, not as apparent as, as it is now, is like everybody needs to have the word agent in their head and be thinking about how to, you know, how to contextualize and integrate everything into that term.
- [48:52]
C
Yeah, yeah, I, I'd say think agent first and then human next. At least that's the mantra that I'm following here at Source, and that's been helping.
- [49:03]
A
Yeah.
- [49:03]
C
John, you were saying something.
- [49:05]
A
I just want people to try an agent today. Right? So go to ChatGPT, go to Claude, go to Manus. All these have agentic flows, have it order, you launch, just see that it can take action in the real world. I think that is a very eye opening thing. And then once you start doing that, and then you can start understanding, you know, what happens if you set up an open claw. And I want to be super clear, the tools right now are not really production grade, right? Like you don't want to be rolling them out across your business, but in three months they will be. In six months they will be. It's going that fast, right? Like the stuff that they're actually capable of doing. So for right now, the takeaway from this episode is get your data foundation layer set up. So sales analytics is a partner of the podcast. I pay for them, Jason pays for them. It is a reasonable fee for them to set up and manage your data warehouse. Now you're not locked in. It is, it is the same infrastructure that every major Fortune 500 company has.
- [49:59]
C
Right.
- [50:00]
A
But they're going to set up, they're going to maintain it, they're going to work really hard to keep your business. There's nothing, I mean, it is, it is blue chip data infrastructure. Then they're going to lock you in because you're going to fall in love with the Sarah's IQ features. The ability to actually query and get answers back about what's happening in your data, your sales every single day set up agentic workflows inside of there, right? Every single day. Tell me how many phone case they sold yesterday. Right? You could do all that right now in Sarah's iq and then eventually they're gonna have agents out there battling the rest of the agents. Krishna, what else do you want to say on the pod?
- [50:32]
B
Let me jump in for a second because you know, when you guys sent out the invite for this one, the reason why I was excited is it says in it leadership and decision making, right? And then, you know, we're talking about leadership and decision making. This is going to relate back to everything that we're talking about here. It's like where does good decision making come from? You know, comes from. I mean the first thing is good data. You know, the, the second thing is, is good judgment, right? And then it's, you know, deci taking decisive action. Like these are, these are like the, so like you gotta have the data foundation to do it. And you know, sure, you can go to Shopify and just like download stuff or look at stuff, but like you just, you have to have this data foundation and then you have to apply, you know, you have to apply judgment. Like data doesn't make the decisions for you, but you just have to have it. And it's all like relates back to leadership because leadership is knowing what matters right in the data and, and particularly like understanding that the data is probably not never going to be totally complete, right? It's like all models are wrong. I say this all the time and, but you have to have like, you get the data as clean as you can. Like, you get it to like Saris is going to get it to like 95% and then you're making this. If you're not making decisions on data, you're just like, you're just using your gut. And honestly that's just like overconfidence, period. Like the level of overconfidence that's out there, it's. There's so many people that are just like, they're like just making decisions on their gut. You know what they're doing, they're, they're like, they're literally, literally flipping a coin.
- [52:26]
A
Sean here to tell you about Saris analytics and Saris Pulse Ridge is profitable every single day. And we've taken that super seriously since we built this business. We track contribution margin by day. We look at the SKUs we sell every single day. And we have to do this manually up until Sarah's launch came out. We take all of our SKU level data, we build it into the data warehouse. Everything that goes into making a true P and L I get on a day to day basis. Sara's pulse gives you clarity. So your CLO and your CFO and your CMO start speaking the same language. Contribution margin shifts teams away from hoping profits survive the season to manage them in real time. Book a walkthrough with The Saris Pulse team today. Click the link in the description. And thank you Saris for bringing you this show.
- [53:08]
C
So I'm, I'm curious actually from a leadership standpoint, how are you getting your company to think AI? Leverage AI? What are some strategies that you are adopting? I'd be interested to know.
- [53:20]
A
Yeah, how do you get buy in across an organization for AI? It's a great question. You know, Jason brought up Craig Foals from Chat Walrus. So Craig, friend of the pod, he probably has AI operators coming out. He's going to be, you know, in the sphere and ecosystem. He was at Crocs setting up AI for them like in the 2020s, right. So like before AI was, was where it is today. So he's seen it from the ground floor. So he has like a training course thing called Chat Waller. So that's the first thing is get it for everybody because we have people who, from 18 years old to 75 years old who work at Ridge. They're going to have different levels of proficiency and you have to teach them what is an LLM, what, what is Claude? How do you log in the basics, right? So make everybody take that course just to, to start off the rip. And then the second most powerful thing is showing what can be done. So we, we've did this in the past. We did game building days where you know, you put teams of five people and you're like, hey, we're going to use Claude. We're going to build the game in an hour, right? And the game is actually fun and engaging and pretty, pretty exciting. Which you can do in an hour in cloud, right? And then, then you show them a demo of what can be done. Like, you know, that's not, that's not a game today. After this I'm gonna do standup. I'm gonna show everybody the cool Manus agentic workflow thing I did. And I'm like, hey, check this out. I post the creative I make in AI all the time. I make amazing ads in AI and I'm posting, I'm like, hey guys, this is what can be done. And then once people are, they have to be, you know, interested. They have to opt in to wanting to learn this stuff. And then we have a bunch of resources. You know, we have an amazing VP of internal projects, Adam who's like the best with AI. We have a guy Jules who's amazing with AI and they're setting up NAN workflows and they're very much like, they are super nerds on the cutting Edge of what's happening. So they're there as a resource, but, like, you have to get people excited to actually try this stuff. So that's how we're getting people to opt in.
- [55:14]
C
So are you going to force people's hand by setting up, let's say something like a AI demo day and say, okay, bring your best ideas and demos, or are you going to let people opt in? I'm asking more like a founder, CEO or the head of the company how you're thinking about it.
- [55:31]
A
Yeah. So we don't start with the AI demo day. What we start with is we're going to build a game. So, like, everyone has to build a game. You're forced to do that, but that should be fun. It's like. And, you know, we do. We do a couple hours of this every, every month, and the winner gets $1,000 or whatever if you built the funnest game. And then people are like, oh, there's money on the line. And, like, just get people's hands in the tool to build a solitaire competitor or whatever. Right. With like, I built Minesweeper, but, like, it's random or something. Like, you just build these little games or whatever, you get people excited. And then it's like, hey, you know, this is something we did. Like, bring us your least favorite thing to do. And then on standup in real time, we're going to build this an AI solution to solve it for you. And we had a guy, Antonio, he is in charge of, like, through a matching inventory ship. So, like, what showed up at the warehouse? What the vendor say? They left. What did the doc say they had? Like, does it all make sense? Is all the costs. That job sucks. Right. We built AI to do all of that, and we did that live in, stand up, and it just shows, like, oh, cool. I don't. I can replace parts of my job that are bad. So that's how. That's how we get people excited about trying.
- [56:38]
B
That is awesome. Yeah. I'm just gonna say, I mean, that's. That's totally the way to do it.
- [56:46]
C
So you're literally gamifying it. Right. Both ways. Right. One is incentivizing people to build stuff, but you're also gamifying it because you're asking them to build games and get excited and start using the tools. So. Great advice.
- [57:00]
A
Yeah. I think that's what high school and middle school and elementary school is going to look like in the future is people building stuff in AI. Jason, you said something earlier about, you know, the data Project's never done, right? And it's because if the data project was done, your business was over, right? Like the, the data is ongoing. It's a box that keeps building and building and building, right? It's never ending skyscraper. Because as you grow every day you have more data coming in every month you have new products launching. And we should just talk about that. That like this is not. There's, there's going to be ongoing maintenance, right? This is, this is your entire livelihood, everything that's ever happened. Your brain is a very powerful agent and you've worked on this business exclusively for five or ten years or whatever. You just have a lot of decisions and data you have to unpack. And teaching that to somebody else so they can write it down and put it into tableau and code and everything else is just, it's, it is a journey. So, you know, this will take a couple months to get started to build your, your, your data warehouse dream. Then every day after that there is a little bit of maintenance going on to it, right? But the power when you finish it is you end up getting something that tells you contribution margin every day, right? Jason, if your marketing team comes to you and they're like, hey, we want to run a 30% off sale, right? Hey, we're 30% off sale. We think revenue is going to go up like this. They never think about contribution margin. But now you can be like, well, I'm just going to ask Sarah's iq, what happens to my contribution margin if we run that 30% off sale? So it is. You unlock a next level of thinking. But work does have to go into it. So Jason, you want to, you want to unpack anything.
- [58:40]
B
Like we're, when we're kind of defining like what is the next good offer, right? That we're going to run for some offer, period. We've got lots of different options, right? We've got gift for purchase, we've got Buy More, save more. There's like all. Or do we just like go with our best discount, our best bundles? Do we just do across the board discount? Like we can, we can actually look at the data from, from previous sales periods, right, and do that. Like, how are you going to do that properly without a data warehouse, you know? And like you got to load all that data in there, right? That's, that's actually, that's dirty work. That's like dirty hard work. That, that's why you need someone like Sarah's to do it, right? It's just like, that's really not aiable problem. It's that's. It's the foundational data layer that every real business. You are not a real business unless you have a foundational data layer. And it does take time to set it up. So this is. There's a little bit of a process here of just having a little patience at the beginning to get it all going. It's not like, oh, I flip on Shopify and I start seeing numbers come. Come through.
- [59:53]
A
Well, dude, you don't get clean data unless you label it. So what you're talking about is labeling the data to actually to make sense and ingest it. But, Krishna, final words for the podcast, my man.
- [60:07]
C
Sorry, Sean, I wasn't ready for that. But final words for the.
- [60:11]
B
I love that. Yeah.
- [60:13]
C
Final words for the podcast. Get your AI foundation ready today, because agents are coming. You will see a serious, serious unlock in productivity. So don't let your team struggle with spreadsheets. Get them on a foundation. Give them their time back. Unleash their creativity to help grow your business.
- [60:35]
A
You heard it here first, guys. The agents are coming. The agents are coming. Krishna, I appreciate you coming on the pod. Thank you for all you do. Thank you for being a proud operator supporter. Jason, great seeing you on the pod, brother. Anything you need from me, you hit me up. We're here. Keep rocking. Talk to you guys later.
- [60:51]
C
See ya. Nice talking to you, both of you.