Intelo.ai Revenue (2025)

Nathan Latka (00:01) Hey folks, my guest today is Jeff Fish. He's a retail technology executive with 20 years experience in enterprise tech, much of it focused on the Chinese market. He co-founded the WeChat management platform called Chatley in 2015, which was acquired by Salesforce in about 2020. He then took a leadership role at Salesforce and helped bring Salesforce on Alibaba Cloud to market. He's now co-founded and co he's now co-founder and co-CEO of Intello.ai, running business operations while his co-founder Rupesh leads product. and engineering. All right, Jeff, you ready to take us the top? Jeff Fish (00:32) Sounds good, Nathan. Nathan Latka (00:33) All right, what are you guys selling? Tell us what Intello is. Jeff Fish (00:36) Merchandising AI for retail merchandisers and planners, everything from MFP to pricing and promotion. And we're an agentic workforce across every every aspect of merchandising and planning. Nathan Latka (00:48) Make that really real for us for a second. You less you list Balenciaga here as a customer. So if I go to Balenciaga's website, wh where are you working in the background? What are you doing? Jeff Fish (00:56) So if you want to get inventory from one store to the other, you're using our agents in the background with their merchandisers and planners to decide what's what's the best store to get the right bag, what's the best store to get the right pair of sneakers, and it will make those decisions and make and drive the intelligence much faster, much more efficiently, and more importantly, much better than than systems that they were using in the past. Nathan Latka (01:18) So so Balenciaga sort of home has produced, you know, La City Backpack Mini. I'm sure I'm butchering that name. You know, they have a hundred, you know, they have a thousand of them that are gonna retail for two thousand four hundred a pop. What you're saying is you'll help the planners at corporate decide which stores around the country to split up those a thousand bags to. Jeff Fish (01:33) Not not just where they should go, but where they should consolidate to. So in the in their example, you know, if you lose that two thousand four hundred ninety dollar sale when a when a client walks in the door, you may lose them forever. And our intelligence is gonna is gonna tell Valenciaga what's the right store to have the right that bag in so that when that client or or that customer walks in the door that the bag will be there. Nathan Latka (01:57) So so how do you do this, right? I mean, that's a hard, hard challenge. I mean, I rarely shop in stores anymore, but it I when I used to, it was so frustrating going in somewhere and saying, I want to buy another one of these dress shirts at Nordstrom. What do you mean you don't have a thirty-six inch, you know, chest that I can buy and then I I never go buy it again? Right. I mean, how do you what data set are you using to train your models to help these brands get more revenue? Jeff Fish (02:17) Yes, that that's a that that's a key challenge for for retailers around the world. And there's a combination of a deterrent deterministic model. So looking at historical sales, looking at the last two years with trend analysis, then looking at the last four to six weeks across all the doors. And then there's an agentic model that allows you to look at, you know, third additional data sets like weather, like physical events, you know, what's going on in the US with World Cup. That would that would drive more footfall for any stores that are in Inner or near the World Cup events. And then economic data, you know, things like spikes in oil causes challenges to different parts of the world. All of those fit into the data set. And then it makes a recommendation. And then because they're agents, it will deliver out a recommendation across any one of our eight agent outputs. But then the merchandisers and planners can then interact with that data. They can question it. They can add additional data sets to it. So they could add, you know, marketing information, they could add performance information all into the data set and then you can get to a final answer and then you can hit execute and then it pushes that information downstream to their warehouse management system or whatever system they're using to to move inventory. Nathan Latka (03:27) And so just give me an example. How would a spike of the per barrel price of oil above a hundred correlate to a decrease in demand for Balenciaga black leather bags in London? Jeff Fish (03:39) Balenciaga is probably not the best example because maybe oil prices won't affect on a weekly basis what what the what the traffic would be. But if you're a high volume retailer and and gas prices go up, that generally will lead to less sales in the stores. So you go to areas where there's more affluent areas where that would less affect them. that that's an example. another example is, you know, shipping's not not doing well in in the Middle East. So maybe your stores in the Middle East, even the the high high net worth individuals are buying less. So you move less inventory into the Middle East because there's less revenue. So there's a pullback in demand, pull back in sales. Nathan Latka (04:18) Interesting. Usually when I interview a founder and they tell a story like what you're telling, right? You sit on a proprietary data set that your crump companies like Balenciaga, they give you access to their metadata, et cetera. Usually I can find some page on their website that lists like a bunch of vertical specific integrations. I don't see that though on your website. What am I missing? Jeff Fish (04:35) Yeah, so we integrate with every data lakehouse architecture. So we're we're partners. We just we just want to startup of the year with with Databricks. So we're built on Databricks and Azure, but we can you know zero copy into Snowflake, zero copy into BigQuery, really really any Lake House architecture that you can find. We zero copy into them. we have a partnership with Microsoft, we have a soon soon to have a partnership with Salesforce, we'll be in the Agent Force partner Network. So those are the key partnerships that we have. Nathan Latka (04:46) Okay. Jeff Fish (05:05) J jet definitely though, our our key our key data set is coming out of any sort of lake house architecture that any retailer would have. Nathan Latka (05:12) So you're skipping all the iPass tools, you're just going directly right to the database. Jeff Fish (05:16) Eighty percent of the time. Eighty percent of the time. Now there are there are cases where, you know, your your retailer's not gone through transformation, their data's sitting in an old instance of SAP or an old instance of Oracle, then we'll build the integration one off or we'll work with an SI partner to do it. Nathan Latka (05:18) Yeah, that makes sense. I see. Okay. Talk to me about money. We love the good stuff here. How do you charge customers? How are you guys making money? Jeff Fish (05:38) We sell it like labor. So we have five agent teams. An agent team comes in, it's deployed within you know 30 to 90 days. pricing can be anywhere from 150 to 450K a year. that that agent team comes in and augments the the teams that we're working with, the physical teams and deli delivers much better outputs. Nathan Latka (06:00) So just to be clear, so if I was like, if I was Balenciaga signing a deal with you and you send over like all the things I could buy from you, I'm gonna see, hey, you could pay, you know, 150 grand for the merchant analytics agent team or 150 grand a year for the MFP agent team. Is that sort of how it's structured? Okay. And and and how do you drive expansion revenue once a customer gets addicted to you? You're providing so much value, they say, Jeff, I want more of your MFP product. Give me more agent, give me more bots on that agent team. Do you sort of charge per seat like per digital seat? Jeff Fish (06:14) That's correct. No, it's not per se. So so it's all in consumption based on the team. They can buy more teams, which is what what we've seen across our our early customers, where they've come in and maybe deployed one or two agents within that team. And then they've added agents within the team itself, and then they've added more teams. So our earliest customers No, no, it's not. so if let's say for example, you start out with assortment planning, which is kind of preseason planning. Nathan Latka (06:47) If that's Balenciaga calling, you better take that call. okay, just Jeff Fish (06:57) the the natural progression would be then you'd go to in-season management. So that would be another agent team. And then you'd take a step back and go, let's build out our merchandising financial plan. So that's an area where we have expansion. Another area where we have expansion is cross-selling. So a lot of retailers will have a separate data set for Europe, separate data set for North America, and a separate data set for APAC. Each one of those data sets would need its own agent team if they're not consolidated. Nathan Latka (07:23) Interesting. Now do you have you guys been around long enough where you can look at net dollar retention over multiple years? Jeff Fish (07:28) So we launched early last year. We've had zero churn. We have twelve customers live on the platform. We're over two million ARR. we've had two of those twelve upsell themselves in the first 12 months. So so far, so good. Nathan Latka (07:42) That's great. Okay. So twenty twenty six is off to a good start. What did you guys finish twenty twenty five at in terms of revenue run rate? Jeff Fish (07:48) Two hundred and fifty K at the end of twenty twenty five, almost a million Q one and over a million in Q two. Nathan Latka (07:55) Okay, and now you're at two million of ARR, meaning last month you did about a hundred and seventy grand of sort of accrual based MRR. Jeff Fish (08:00) No, no, no, no, no, no. We did quarter million total last year, which is our first year generally available. Then we added almost a million in ARR in Q one. Then we added an additional million plus in Q two. So we're over two million total. Nathan Latka (08:14) Yeah, so two million ARR divided by twelve months would be $160,000 a month. That's what I was referring to. So last month. Yeah, yeah. So you guys are last month you did more than $160,000 of monthly revenue. Interesting. Tell me, these are well, relatively large contracts, right? Anyone listening can take two million of ARR divided by 12 customers. How did you get your first five customers? Jeff Fish (08:19) Right. yeah, correct. That's correct. Founder led sales. Yeah. So Rupash, my co-founder and co CEO, myself, and we have another gentleman in in Europe, Vasilis, who came out of luxury. So we're focused on luxury and specialty. The first five were very much founder led. Now we have two enterprise sellers. We've got consultants that are working with us, and we're we're growing pretty rapidly now. Nathan Latka (08:56) Yeah, I was gonna say r respectfully, Jeff, you don't look like quite the fashionista when I look at your background. So I was wondering where the the luxury designer gene came in here. Jeff Fish (09:06) Yeah. you know, I when I started Chatley back in twenty fifteen, a lot of our customers were luxury because it was China. And you build those relationships over time. And then if you then fast forward to after we got acquired by Salesforce, just because of the nature of the business in China, most of our early customers were carings of the world, L VMHs of the world, Richimon's, and customer zero for us was was Gucci on in caring. And when I decided to do this with Rupesh, the first call I made was to our customer at at Gucci and convinced him to to come over and and do this with us. Nathan Latka (09:46) What so so you basically that Gucci the reason you got Gucci as your first customer is you had a relationship with whoever the buyer was there from your Chatley days and you s that sort of followed you through the past, you know, four or five, six years. Jeff Fish (09:56) No, so Gu Gucci was an early adopter in China through Salesforce. They weren't a Chatly customer. Obviously, Salesforce has a has a large a large array of customers across all verticals, but the the the need and demand in China was really luxury first for compliance reasons, for performance reasons. So L VMH and Kering were the first customers in China for Salesforce. Nathan Latka (10:18) Very cool. And then how did you and Rupesh meet? Were you at? Let me see here in the background. Was it Chatley or no? I don't see anything. Where was the overlap? Jeff Fish (10:25) that I worked for him at Symphony Talent. So before before I started Chatley, he he was he was CEO and I I was had a product for a while. Nathan Latka (10:36) I see. Okay, so you get back together and are so is just you two co founders? Jeff Fish (10:40) We're the two main co-founders, yeah. Nathan Latka (10:42) You guys just play nicely, you fifty fifty. Jeff Fish (10:45) We do. Yeah, yeah. He he runs product in engineering. I run the business. Nathan Latka (10:46) Okay, that's that's nice. Okay, and tell me more about how you've capitalized the business. Bootstrap today or have you raised? Jeff Fish (10:54) We bootstrapped it at first. we did a pre seed venture around with Illuminate Ventures. So Illuminate's on our cap table, and we're in the process of raising right now. Nathan Latka (11:04) Okay, so that was a pre seed last year in 2025 for two million in October. Can you share what you saw? I mean, did you do a price drown or was it a safe or a note or convertible note? okay. Okay. Well d what were you s when you went out in the market, right? You know, everyone's wondering, hey, how are investors, especially pre seed pricing these AI startups, especially with two founders with a really, you know, great background? Are you I mean, are you able to get up to like, you know, I don't know, a hundred crazy hundred million pre seed numbers, or is it pretty sort of standard, you know, ten million cap sort of thing? Jeff Fish (11:08) In October, yeah. We did a safe. It w it was in the ten million range. It wasn't in the hundred million range. Nathan Latka (11:37) Okay. And what are you seeing now? You said you're looking at me potentially doing something now. You've grown dramatically over the past eighteen months. What's the market telling you? Jeff Fish (11:45) So the market's telling us that there's definitely demand. So we we've already, you know, received our first offers and term sheets. we're still negotiating those. I think that the the the demand and and and the pre seed valuations are always a negotiation, right? but they're they're well in line with where we'd expect. And, you know, we're we're have a clear path for this year to be over four million ARR. And you know, that's what that's what we're valuing the company at. Nathan Latka (12:15) What are you how have you structured the team today? Do you have quota carrying AEs yet, or are still you and Rupesh closing the big deals? Jeff Fish (12:21) No, we have two now. Yeah, yeah. So in March we we recruited someone from Salesforce to to run Italy and Switzerland. And then it we also at the same time recruited someone from Autone, which is a competitor in France for merchandising and planning. She was the she was the country manager for France. And now she's our country manager for France, Spain and and and UK. And you know, they they both started in March. They're they're already past their first deal and working on their pipe and gonna have two to three deals by by the end of this quarter. Nathan Latka (12:52) Why'd you decide to use human AEs when we're seeing all this buzz right now about just hire these, you know, you know, artisan will do it for you. Just install them on your website. Digital AEs can close deals. Jeff Fish (13:03) we have digital SDRs. I mean, we we definitely have built, you know, we're an agentic company, so we built our own agents for for SDRs. But when you're, you know, selling to to the Chanels and Balenciagas Dolce Gabbanas of the world, or, you know, high volume retailers here in the US like Express or Children's Place, you're you're not doing that with with a with a agent. You have to have a physical person, go have a meeting and build a relationship and expand that relationship and you know, that that's not going away anytime soon. Nathan Latka (13:31) So founder led sales to two AEs hired here in March as you guys are breaking in a one to two million of revenue. What's the total team size today? Twenty eight. Okay, interesting. Okay, so twenty eight folks, I imagine what how many are engineering? Jeff Fish (13:38) twenty eight. the core engineers, we've got about seven. And we're gonna this is the the beauty of today versus when I started my company over 10 years ago. We'll max out at 10. we we've you know, we're already at 90% code generation. we're planning to get to 100% code generation by the end of the year. So we kind of have this new type of role architect, product, engineering all together. that's that'll max out at 10. Then we've got integration engineers, which, you know, to your point earlier around integrating with legacy systems like. Like an SAP or Oracle, we need those and then implementation and and customer success. Nathan Latka (14:20) And if you, you know, you're obviously testing the market right in terms of raising capital, but even at Founder Path, you know, as we're investing in companies on the debt side, we always try and find what their AI moat is. So we're looking for is there a proprietary data moat? Are the founders just super connected? There's a Rolodex moat. Is there some other, you know, ETL process they run that has alpha? That could be a moat in the age of AI. What what do you think your strongest moat is? Jeff Fish (14:41) Think there's two. I I I think our merchant analytics and the context that every one of our agents is gaining with every customer that we onboard is a moat in and of itself. So, you know, when we launched our first customer over a year ago, we we joke around internally. It was like launching a junior allocator. And we get their data, we train the model, and on day one, it's it's it's good, but it needs a lot of time to get to get better and smarter. Now that we've got you know over 10 customers across, you know. a pretty wide range of apparel and fashion and and specialty. The day that we turn on any one of our agents for a customer, it's like they've hired a senior director at that level. And it's just getting smarter and better over time. And that's our mode against against our competitors. So you know what we're what we're playing in is legacy software, kind of more modern black box AI, and then you know the CIO that goes, well, why can't I just do this with Claude? And the The legacy software is easy, right? We're much more nimble, we're much more modern, we can deliver much faster with better outcomes. The the more modern systems that we compete against that are more black box AI, we have the explainability, we're a system of action. And then the CIO that goes, why can't I just do this with Claude? They don't have the context, right? We've got the context that's already preloaded and is getting better over time, every single agent deployment we have across our customer base. So that's our moat on the capabilities of the agents. And then in the space that we're in, it's very much a snowball effect. So we started with luxury and specialty because our the founders knew that space really well. And it's a fast follow business. So now when we show up to an event in Paris, it's a who's who of of brands that want to come see hear what our customers have to say. And that's the case in New York. That's the case in Milan, and that's continuing. So it is the Rolodex piece. It's also the the true technology piece of how we built out the Asian architect. Nathan Latka (16:36) Can you just teach me what you say when you mean training your model? A lot of people are throwing these buzzwords around and it's it's nothing more than a glorified Excel file, right? So what do you when you say training your model? Jeff Fish (16:45) Yeah, so the data, the data that comes into the agent. So the data that's coming from the customer. So historical sales, historical inventory across all of our customers. It's the the prompts and and the conversational UI that are that are happening at every customer tenant level. And then it's also the interactions with that data at every customer tenant level. That trains the model across the board. Now, obviously, we don't share, you know, one Balenciaga's customer data with Dolce Gabbana's customer data, but the context that's happening within within the the conversational UI, the prompts that are happening, that all feeds up to Intello and that makes the agent smarter every time we we have a neck next deployment. Nathan Latka (17:22) We all saw sort of Alex Carp go crazy on CNBC. We see thinking machines launching basically, let's just for the sake, an on-prem version of a foundation model. So you keep your alpha, you keep your intelligence. W if I'm Balenciaga signing up to Intello and feeding you all of my data and you find some alpha there, right? You find a signal in the noise. Maybe that signal is, Hey, I'm making this up, right? When oil prices go above a hundred dollars, you should ship more to your stores in New York. Right, that's maybe real alpha Balenciaga has. You then anonymize that and put it in your master model that then will get applied and help Dolce and Gabbana grow their sales in the future. How how do you on a sales call sort of counter that argument from a Dolce and Gabbana or a Balenciaga? Jeff Fish (18:06) Yeah. So the way that we counter it is they're currently using systems, whether they be legacy systems or doing it in Excel or or doing it themselves that aren't scaling for them. They're not getting the outcomes that they need. They they're especially in retail, they're never going to have more opacts. They're not able to add more bodies to get a better outcome. With our agents, literally in 30 days they see ROI. In 90 days, they see significant increase in sell through. And yes, other customers will see that too. But their business model is gonna be very different. And their business itself is gonna be very different than every other business. And in in merchandising and planning, which is the nervous system of retail, there's gonna be that art and that creativity and that gut feeling around the brand that an agent or an AI of any sort will never be able to deliver. And we marry those things together. And that's that's the that's the value that our customers are getting out of out of our agents. Nathan Latka (18:58) And with that wedge, how do you steal market share from the following three competitors? These are the folks my research team sent over. I'm preparing for this interview. They said, Nathan, Blue Yonder the 800-pound gorilla in this space owned by Panasonic, 8.5 billion valuation in 2020 twenty-one over a billion of revenue. That's a legacy system that you effectively want to convince your customers to rip out, I think. O nine solutions would be like the digital brain planning platform they raised at a three point seven billion valuation. I guess they're big in merchandising and and demand planning for enterprise retail. And then lastly you've got the legacy folks you've already talked about, Oracle retail and SAP, which you want to go displace. Is your wedge against all these folks just you're moving much faster and you're agentic first? Jeff Fish (19:37) More than that. let let's let's use the first thing you talked about. we have customers that have them both. And we become the intelligence layer on top of it. So we'll come in and go, look, in 30 days, you're gonna get better outcomes with those systems that you have that you're paying three to five million a year for or more, right? Just using our intelligence layer. And then that's why you need humans working in in sales and customer success. Then we can come in and show them over time. Well, yet 90% of the work that's being done at that intelligence layer is being done by Intello. Why spend $2 million on this module from a Blue Yonder or from an 09? Just come in and replace that over time. And that's, you know, that's a two-year cell cycle because generally those contracts are pretty long, pretty much longer. We'll come in and just replace those, but we'll do it piece by piece. So we'll deploy a set of agents or an agent team for in-season and then replace that piece over time, then do assortment planning, which is a lot of art and a lot more art than science. You know, think about the creativity of. I think a red shirt's gonna do better next season than a blue shirt. Well, why? Here's all the data behind it that tells us this. So these are the capabilities that that we've got there. Nathan Latka (20:40) And and out of curiosity, I mean, you know, an agentic workforce sounds like they're using a lot of credits per month. On your hundred and sixty thousand dollars per month on revenue, how much are you then turning around and just spending directly to open eye or the foundation models in terms of your credit cogs? Jeff Fish (20:54) Yeah, you know, it's a great question. Something we review every week. we are not burning through tons of tokens at the agent level. because we have the deterministic model on top and and we're running the ML model before we even get to the tokens, that's that's covering 80 to 90 percent of of the runs that we're doing with our customers. actually, it's I I I I was a bit shocked when we were reviewing this at the end of last year on how few tokens we're burning through because of our deterministic models. That doesn't mean we don't have big invert costs. We're we're consuming massive amounts of data, especially in our Databricks architecture. But it's it's not going it's not going to the the frontier models. Nathan Latka (21:33) Yep, very cool. Well, Jeff, anything we should have touched on that I didn't ask you about? Jeff Fish (21:36) I you know, I'd love to talk a little bit with with a minute or two left about what's next. I think what what's next is big. Yeah. we're gonna be rolling out pricing and promotion, which is an agent that, you know, we or agent team that we've just started on. That's gonna get us into, you know, beyond specialty and luxury, big box retail, multi-brand retail, and then we'll move into vendor management, which will which will get us into areas like grocery. So these are areas that we'll we'll tap into. And our you know, five year plan is to really own merchandising and planning. Nathan Latka (21:41) Hit it, hit it. Jeff Fish (22:05) In every region of the world in in all in all types of retail. So that that that that's our expectation. Nathan Latka (22:10) Is it fair to say with these additional product ads, you think more of your growth over the next twelve to twenty four months is gonna come from getting more wallet share from your current customers, or it's gonna come from net new customer ads? Jeff Fish (22:20) The next twelve to twenty-four months, we're gonna stay in our lane, which is right now specialty and luxury retail. get more wallet share from all of them. So we think that there's increase from all of our install base, but we also think we're gonna significantly increase our our customer count over the next two to three years. Nathan Latka (22:37) On that note, Jeff, if people want to follow your story after this interview, where can they find you online? Jeff Fish (22:41) they can find me on LinkedIn, on on X, on Intello.ai, lots of places to find me. Nathan Latka (22:47) Guys, he's been there and done that. Sold his first company to Salesforce, cut his teeth in e-commerce and the Chinese marketplace. Eventually in 2025, they launched Intello.ai with his buddy in the space. They were nice. They split equity 50-50, got a pre-seed round done for $2 million, around a $10 million cap to really hit the races. First year they finished with $250,000 of ARR. That was December 2025. Fast forward here to today, in 2026, they're currently doing about a $2 million. Annual recurring revenue run rate across 12 customers. They're now scaling their team, hire their first two AEs in March. Before that, it was pure founder led sales. Now 28 folks full time, seven engineers. And what they're doing is again they're forward-deploying AI into these massive retail brands like Balenciaga, like Dolce and Gabbana to help them plan. Hey, we have all millions of SKUs. Where do we set the inventory? How do we increase pricing? How do we decrease costs? And he's hoping he can effectively use his agent armies to replace their physical workforces or at least augment. them and make them even smarter. Jeff, thanks for taking us to the top. Jeff Fish (23:44) Thanks, Nathan. Take care. Nathan Latka (23:45) All right. All right guys. Cut Jeff, what do you think? You have fun? Jeff Fish (23:48) Yeah, it was great. How about you? What do you think? You you do this all day long. Nathan Latka (23:50) Cool. I appreciate you making time for me. Well g I thought it was awesome. I thought it was awesome. I was awesome. The more transparent, the more vulnerability. Like that's that's why I do this show. It's so much fun. So it sounds like you listen. Is there anything you want to tell me to do differently on more, you know, future interviews so you get more value listening? Jeff Fish (24:06) No, the questions are great. I didn't expect you to jump around on the website. I I probably would have directed you on where to go. But except for that, that was great. Nathan Latka (24:14) If you have B-roll of a demo account inside of your app that you want me to build in with my post-production team, send that over to me. Like a screen share or something. All right, see ya. Jeff Fish (24:21) Yeah, no problem. I'll send you that. Thanks, David. Take