DEPLOY STAGE

Our program is going to begin in a few minutes. It's time to find your seat. As a courtesy to our speakers and fellow attendees, please take a moment and silence your devices. Let's give a warm welcome to the Chief Revenue Officer from Shoplaza, Adam Maudsley.

Mornin' all!

Hope everyone's doing well. Last day of SaaSTA. It's looking like I'm pulling in a big crowd, but we'll get through it anyway. So yeah, I'm Chief Revenue Officer of Shoplaza and Subotis, and here to talk about from prompt to revenue, how AI agents are rewriting the commerce stack and talk through some of our experience that we've seen. So first of all, we were going to do a live demo, but from the reliability of the Wi-Fi, we've pivoted and we'll show a little bit of a video after this instead of trusting the Wi-Fi. What we've seen currently, what used to take weeks can only take seconds or days or minutes nowadays. We have a whole heap of agents and technology on our stack that we have helped make our experience and our technology scalable, fast, and effective. So the example here, which we'll see in the video, is now you can create a new business and a new website just from a prompt and a sentence. Previously, you used to see commerce platforms were very complicated, huge amount of point solutions, different pieces that you'd have to get up and running. Whereas now, obviously, with AI and agents, the theme of this is making it as easy as possible. So instead of scaling businesses, people used to have to scale dashboards and have more and more different tools and systems. What Shoplaza has done is trying to curate that and make it all less complicated. So the following video, fingers and toes crossed. Okay, I've killed the sound because you're not here just to watch a video, but this is an example. You can go to Shoplaza, basically write a sentence on what you want, it creates a website for you, and then we have about 7 agents plus more and more rolling out that can help automate and make it as easy as possible. So image creation, we have agents that can create professional web assets content. Then also, I like to use the Field of Dreams, if you build it, will they come? Not necessarily. So we also have agents and a Lavasa studio that helps you with AdWords advertising campaigns, return on ad spend, because it's not that easy nowadays to get yourself found in a very busy marketplace. Then we have payments. We've been processing billions of dollars over 650,000 merchants. So tax risk compliance, making it all easier for you. And then of course, the usual agent stuff for support and operations as well. So that's some of the agents, and then I'm going to dive into a little bit more details. So what we've found is the market is definitely changing. AI is here to stay. That's not a change, and I love the SaaS-pocalypse. I don't know if you saw the dancing panda yesterday and even SaaS is dead, whatever you call those flying billboards behind the planes. Our take, it's not that. I think one thing that AI really makes possible now is more data. Whether it's workflows or content or platforms. I'm on the Vercel stage. You've got Lovable, you've got Claude, you've got all of these tools. Everyone has access to the same tools nowadays. But that is not really necessarily going to make you successful. What makes it successful is the data and making it all work together. So from our experience, especially in the commerce platform space, we've seen traditionally, I think they say about 24 different systems within a commerce stack, and that can become complicated. So I think the 60% of projects failing, that was a Gartner study. Which is actually an improvement. Last year, I think the year before, MIT study said 90% are failing. So it's an improvement. But why do they fail? Well, I think there's a few reasons that we have seen. Integration complexity is one. Again, you say 24 different systems within a commerce stack on average. But Shopify, amazing platform. But I did look, they had 800 new apps launched last week alone, and now they've got about 18,000 apps. So really important and powerful if you know what you're doing. But not everyone's an expert on every single thing. So we're finding that making sure that you can actually curate that and pull it together is really important because in the modern world with AI and everything, to be honest, I've had some great time walking around talking to sponsors, but some of the times I'm like, I don't even understand that. What do you guys do? And I find myself not to be arrogant, but reasonably tech savvy being in the industry. Then I did count a number of billboards along the 101 from San Francisco this morning and I'm like, "Don't know what you do, don't know what you do, don't know what you do." Expecting your small customers to understand all of that is not necessarily going to be the most effective way to do business. Also, the 3.2 more dashboards, I guess from our experience, we think curation and content data is actually really important. I've heard a few people say data is the new moat.

We agree.

But making sure that you actually use it in a comprehensive way is very important because I don't know about you, even with Pass IDs and key passes and things like that, if I have to save another password and log into another dashboard, it's not going to help my business. So from our point of view, we see that the data and the context is really key for us. And that's why we've used a whole heap of agents for ourselves to to make sure that it's making it as simple as possible. Because as I said, we're not really hearing that everyone's asking for more AI. It's more, what are you actually using it for? So come in with generic context and a generic prompt, you're going to get generic results. So be specific and make sure that if you can, not just 24 different integrations. MCPs have been awesome making things talk together, but does that really actually pull in a continuous— continuous? Is that how I say it? Continuous data layer that actually has a really comprehensive vision across the stack. So for our take, we say commerce doesn't need more AI magic. There's plenty of AI out there and lots of tools and amazing stuff. But for us, what we think is important is you need to actually have AI and agents that connect all of your details. So execution and actually having real outcomes instead of just creating more data and giving more workflows. So that's what we've found really works for us. I know that I've heard a few people think point solutions is the key. I guess our take would probably be a little bit counter. Point solutions are awesome if you really know what you're doing, but there's a lot of things out there. So actually curating what you give people is actually a very powerful tool. Let people actually be successful and focus on their business so they don't actually have to run around and look at, how am I an expert in 50 different platforms or 100 different apps and 100 different systems? So Shoplaza, I think, not sure if anyone, take a few minutes to explain what Shoplaza is. Couple of years ago, so Shoplaza, by the way, e-commerce or commerce platform has 650,000 merchants doing billions of dollars. So very successful, but we made the conscious decision a couple of years ago to say, let's rebuild the platform and actually make it AI, I guess not native, AI first. Because we found that a lot of SaaS companies, you talk about SaaS apocalypse, if you're just bolting AI on, you potentially become a feature and you become a tool of one of those 24+ systems in someone else's record and someone else's platform. So what we did, we rebuilt everything, and basically what came out of rebuilding that is what we call the Shoplaza AI Commerce OS. So basically 7 live agents at the moment, but a lot more coming. And I already mentioned it in the video, but easy way to use an agent to create a store, professional images and ad creation, ads return on services, avoid having to pay for an agency— no offense if if anyone's from an agency here. Athena, which is operations. There's obviously payments, risk, support on top of that as well. So what we've done is pull those all together and make it a system that actually talks across our whole stack, which has billions of dollars. So when we see improvements, we don't just roll it out to one customer, we roll it out to all of our customers. So I think the compounding nature of having all of that data and all of those systems talking to each other It's not just an MCP endpoint that basically has one system of record here and another dashboard here. We've centralized that, and that's how we've found for us that agents have really, really had an exponential effect on our business, and not just our business, obviously, the success of our customers, which e-commerce and payments and sales, it's a usage-based platform. So if we're not successful, we don't want to just charge people a high licensing fee. We're all about usage-based and making sure that the agents make them sell more or acquire more customers and focus on their business. So with that, we also have launched Subbotiz. So Subbotiz is taking the skill set that Shoplaza have had over billions of dollars and globally, and we've built an AI SaaS subscription usage platform. So think, obviously, the the plans, the upgrades, the free trials, the usage-based billing. Really important nowadays, because I think everyone can launch a new product in hours, but if you can't commercialize it and monetize it, then you've got a demo. You don't have a product. And one thing I was talking to one SaaS founder, and I'm not going to mention who it was, but they launched and they thought they were doing great. And then at the end of 3 months, their AI token deal ran out and they got the bill on the 4th month. And they looked at it and like, oh my God, I've lost money on every single customer. So making sure you have a flexible billing system that's built for this new world is really important, 'cause unless you're one of the lucky few AI companies who have endless funding and can basically not make money, our take is you shouldn't be afraid to make money. And this is what we've built to kind of help you make that step and launch from day one and monetize from day one. So tax, sales, compliance, MCPs, whatever you want. It basically means you don't have to become a fraud, a tax, a payment expert. We basically offboard that. Just like with the Shoplaza toolset, our concept here is we basically take the heavy lifting and the plumbing so you don't have to become an expert. So that is our take on, on Sobotis and why Shoplaza and Sobotis work together. Just in case you're wondering why I'm talking for both. We're all part of one happy ecosystem family. And then I'm already at time, but oh, sorry, they've got a cool running out of time thing flashing up here. So a few takeaways. I think prompts is basically, it's here to stay. If you can make it simple and make prompts and AI agents solve customer problems, I think creating the new workflow and just just basically using it for the sake of it, is a trap you have to avoid. Make sure you're building it for the right reasons. If you rewrite your stack with an ending mind to actually make sure you're helping a customer, or solving a pain point, or removing administration, that is important. And I think number 3, I would say, was probably the most important that we've seen, the shared context. If you're just doing one thing in the point system, everyone wants to be the data layer, or the merchant of record, or the, you know, system ledger, whatever you want to call it. And that is where I think it's really important. If you're just going to become a point solution, there's lots of needs for that. But if your business can actually have context across all of your systems, that's where you can really get AI to help you consolidate and make sure you have improvements at scale. And then last point, fast launches matter. Obviously, from our point of view, you can create a website in minutes, or you can launch your new SaaS subscription billing company, a gong of whatever, insert name here. Not naming names, but there's a lot of companies doing that and great for them. But you can also use Subodhis, of course, final pitch to actually launch or use Shoplaza to create your e-commerce platform and basically take care of the plumbing and make sure you guys are successful in whatever you are supposed to be doing. So final one, magic impresses people once, We put systems compound. Our take, AI is very powerful, but make sure you have the systems talking together. Otherwise you have disparate systems and things that don't necessarily work cohesively. And with that, I'm already over. So we would love to talk to you if you haven't come talk to us already. We're over in the other hangar in, I think, the second avenue, someone called it. I like the avenue instead of the street. So we are Shoplaza and Subotis at the booth. Some of the team are here today, and a huge crowd, so thank you for your patience, and have a good rest of the show.

Please welcome to the stage the Vice President of Product Advocacy from NewU, James MacArthur.

How we doing, everybody? Alright, my name is James MacArthur. I'm the VP of Product Advocacy at NewU.io. Now, the first question I always get is, what does that title actually mean? The answer is, it's kind of whatever I end up doing in a given day. So strategy and operations, customer advocacy, all of that. Now, second question I get is, what actually is new? New is Salesforce-native revenue architecture. What I mean by that is everything from the moment you decide to create a product catalog, all the way through to when you're finally billing your client. Making every single step of that journey easier and simpler for every one of your reps, enabling you to take incredibly complex pricing catalogs, incredibly complex selling motions, and make them simpler, easier, and faster for everyone across the board. Built with the build once, deploy everywhere model, you can put this on your website, in your app, Everything you see me do today, you're able to do through your website. So let me tell you what we're gonna do today. This is all live demo. This is the only slide I got. I'm gonna take you through how new functions from an AI standpoint. What we're doing with AI is one of the coolest things I've ever seen. We're gonna, I'm gonna take you through if we have time, a renewal, what that looks like using both AI and Nu. And then I can take you through some ways we can extend everything, at which point I want questions. I want to be derailed. So to get started, I'm going to jump into Nu. This is our demo environment. Looks like Salesforce. Again, Nu is Salesforce native. What that means We're using your standard opportunity flow. Your reps never leave Salesforce. Your admins control access the same way they always have, using permission sets and profiles. Every single thing you see me do today, though, can be done via API. What that means is everything's headless. Not in a— takes a long time kind of headless. I mean, we've got, we're dual platform, so we have an entire AWS architecture that backs this up where you can have sub-second calls happening in real time. But today we're here for AI, 'cause this is SaaStr AI, and we're gonna go through what this looks like. How many people in this audience have ever had a rep go, I don't wanna create 3 different quotes, I need to create a couple different variations for this, and it takes them about 2 hours to do?

Yeah.

Classic. So today, my rep's gonna say, hey, can you create 3 quotes for me? I need a 1, 2, and 3-year version. Of Vroom Enterprise, which is our enterprise bundle, not news, obviously, it's a made-up company that I don't run. All right, Vroom Enterprise, with all the recommended options. Do 125 units, and give a 5% discount, on each additional year. At this point, my chat isn't working. Claude is in process. It's gonna say, hey, I don't have any opportunities, but here's the pricing. Here's what the bundle will look like. I can now say name it. What you want, close it the third Thursday of next fiscal quarter because you want to do a random date so that your VP of sales— shout out to ours, Travis, back there— who you don't want them to get wise to you. That's never fun. So third Thursday of next fiscal quarter in the proposal phase. Now, Nu is built on an incredible data model. Our data model was built by our CTO and engineering teams. Our CTO was at SteelBrick, Oracle CPQ. She built NeoCRM in China. She built Zora Billing. After all of that, she said, "I hate everything." and built new. The bones of new are incredible. What that means when it comes to our AI solution is that now we're fully deterministic. We can give you consistent outputs every single time. It's going to ask you questions. It's never going to assume it knows what you're trying to do and why. At this point, does this look correct? Go for it. And it's off to the races. It's gonna confirm its understanding, create the opportunity for me with the correct date, and then generate those 3 quotes with the tiered discounts. But it didn't actually generate the quotes, 'cause what we don't need is a bunch of additional data in the system that nobody's actually gonna use. What it did was it previewed those quotes for me. Not only did it preview, it's showing me the additional discount at each level, and it's gonna give me a comparison between all three. So if I'm a rep and I'm on the phone with a customer, you can imagine me going back and saying, Hey, here are the different pricing options. Here's what I have. All of this is pending approval, but I can get those 3 different options, get pricing on the fly, have an opportunity generated, and know exactly what I'm going to be selling this customer before I even get to the stage of creating a functional quote. Incredibly speeding up. That's not a good— Phrase, whatever, it's fine. Speeding up your entire sales cycle because now you're giving transparency to your customers, you're giving flexibility to your reps, but all of this lives within the bounds of your Salesforce pricing engine. So you never— you control how your reps price. You control the bones of what they're doing. They're not— this isn't a Google spreadsheet. You can't change change values. This is going to do all of that for you in an easy and simple way. So now as a rep, I'm like, okay, good, talking to them, they want the Gold Support option, they want to do the 2-year. Great, let's create the 2-year, add in Gold Support instead. Of standard and create. And now that I'm ready to create that quote, I could have changed the discount, I could have changed any of the pricing within the bounds, but New at this point is gonna go ahead, create that 2-year quote with the $10,000 a year of the standard, instead of the standard support. We We have 24 months, at which point I can launch it. Now, this didn't create some quote elsewhere in the ecosystem. It created it in Salesforce. This is the standard quote object, quote line item, conference Wi-Fi. And when I hit go, it's going to take me directly into my quote line editor, where my rep can actually start to play, because I mentioned that the bones of new are great. When it— it's tethered off my phone right now, so we're playing games with the Wi-Fi. But this is where the rep can play. We've created a quote that has implementation, we've got gold support, we have 2 years. You can automatically see this ramping year over year. Out of the box. Now the rep says, you know what, let's do a 16% discount. Now this is highlighted yellow. Why is this highlighted yellow? Because sailing on top of new, we have yet another AI solution, and that is our pricing plugins. So one of the things we get very often is custom product requests. Everybody here You've got products. How many of your customers are sitting there going, I need this, I need that, I need this to show in the UI, I need that to show in the UI? With New, we have our Price Builder chatbot where you can build functional visual approvals, show, put warnings out there for your customers, put a ceiling, a discount ceiling, a discount floor, discount floor. So if my rep then says, you know what, actually, let's do a 76% discount, it's going to cap it at 55%. If I go and say, you know what, okay, we'll do a 6% discount, all warnings are removed. And it does this on a line item by line item level. Designed to be quick, designed to be easy. Now, I can also make changes on individual lines, and it'll automatically calculate. You'll notice how fast it's calculating. That's because sitting behind New is an incredibly powerful pricing engine. This pricing engine not only enables real-time discounting, real-time changes, but most powerful proration engine. For those people that are CPQ nerds, You'll know proration is hard. New can prorate down to a single day or out to a month, all in real time as you type. But that's not all New is. New is an end-to-end sales and billing solution, CPQ all the way through. Full quote to cash. Nothing highlights that more than the fact that I can bill upfront, I can show my customer at the time of quoting every single bill they're ever gonna get. So how often do you have customers go, "You know what, I wanna know what's the first bill gonna look like? What's that first invoice?" With new, I can do it. Not only can I do it quickly, But I can actually show you exactly what the items in each of those bill invoices is going to be. And because we're connected end-to-end, it just doesn't go into some black hole and pop out the other side. You have full capability to control. This is just what you quote is what you're going to bill. Whether it's consumption, professional services, one-time products, recurring products. Everything across the sun. If you can quote it, you can bill it. All in one place, all Salesforce native. Your reps never have to leave. Now, finally, we're ready to go and actually close one, this deal. So rep comes into our opportunity. We could have approvals layered on top of this. We can have a whole bunch of other steps in here. We can do— we do do DocGen. It can be done agentically. We have Approvals Pro, which is an advanced approval system entirely built within Salesforce. But for now, I'm just going to go ahead, hit Close One. Oh, you know what I didn't do? Live demos are fun. I didn't hit Primary. So I'll go ahead and hit Primary Quote. No, I don't want to see notifications. And as you can see, this is Salesforce. So I can do everything there. I can finalize my order, which will generate it on the new side. And this you can think about for your Deal Desk team. Your Deal Desk team wants to review every single order that comes through. We have a lot of customers that'll feed this into a pending, pending billing situation. So from here, I'll hit this and then save and exit, activate, and we're good to go. Now we've got about a minute left, so I'm gonna show you really quick what new subscription management looks like. So behind new, and this is the reason why I came to new, because 3 years ago, I was sitting on a beach with my family. I was living my best life, I was consulting about 20 hours a week, when our CEO showed me this. Every subscription, everything you do is tracked. You have full visibility of the revenue over time for every single one of your customers. You have a customer call you, say, "You know what? I wanna add 25 licenses to this. I wanna do it, yeah, I wanna do it now. I don't need to get a quote in. I don't need to have any of that." I can hit 2 buttons, Express Checkout, and wait for Salesforce to catch up to itself. Refresh, and I've just processed an upsell. I've just processed a change order. So when I go to my orders, I'm going to have two of them. I have my original and my change order. Giving you the full history of Land and Expand in one easy place. Nu's designed to be simple. It's designed to be easy. We went the long way around here today. We can speed this up to where it can be one click, two clicks, any of it. That is my time. You all are amazing. I'm gonna be over there for a little bit, or in booth SG09 all day. Happy to talk through all of this with everyone here. Thank you all very much, and I hope you have an amazing SaaStr.

Sweet. Joining us now from Papaya Global is the Vice President of Client Success, Sivan Fishel, and the Director of Product Experience, Hagit Bensour.

Hi everyone, thank you so much for joining us today. My name is Sivan Fishel. I'm VP Client Success at Papaya Global, and I have with me today Hagit, our Director of Product experience. And today's title is "Your AI Agent Will Get You Sued. Here's How We Fixed Ours." And yes, we mean it. You see, our clients operate in dozens of countries. They have access to our knowledge base, our guides, our experts. But at the end of the day, in the middle of the night, when they have a compliance question that's simple, whether or not they can terminate a worker in Germany, they don't come to our knowledge base. They go to ChatGPT. And ChatGPT can give them an answer that sounds very confident, but is completely wrong. One wrong answer can cost you $250,000 or more. And this got me thinking, how can we protect our clients from getting sued? And how can we create trust that will make them come to our knowledge base as opposed to going to ChatGPT in the middle of the night? We operate in 160 countries. We've reviewed thousands of contracts. We've gone through thousands of terminations. We have years of compliance experience. You can easily call us dinosaurs in the industry. How do we educate our clients to come to our knowledge base? How do we extract the information from our knowledge base and create our own AI agent? And can we put all of that information into the AI agent tool that we built? Hagit led the build, and then we had a very difficult question afterwards: when will we be able to release it to our clients? But I'm gonna park that question for one moment, and I'm gonna let you— I'm gonna let Hagit show you what she built.

Thank you, Sivan. So let me tell you how everything started, because it did not start well. I started very simple. I took a real employment contract, a Brazilian CLT contract, and gave it to Claude. Then the same contract to ChatGPT, same question, document. Claude looked confident, ChatGPT looked also confident, but it gave me completely different answers. Next slide, please. And when I checked both against the actual law, neither one got it fully right. That's when I realized that the problem is not which AI you use, The problem is that nobody taught any of them to think about compliance. Next slide. So I stopped asking which AI is better. I started asking, why are they wrong? And every time I found a mistake, I turn it into a rule. Rule number 4: Don't guess jurisdiction. Rule number 8: Don't sound confident if you are unsure. Rule number 18: Don't flag without citing the right law. One by one, over weeks, 22 rules total. Next slide, please. So here's what that looks in practice. Claude, missed it. It was very confident, professional, but missed weekly hours and just cause issues entirely. ChatGPT cried wolf. It flagged 8 issues. Most weren't real contract problems, just statutory rights. Our agent, Papaya One, got it fully right. 9 clauses reviewed, 2 non-compliant, 7 confirmed compliant, every issue sourced. Next slide. So now you probably asking yourself what was the difference. So the difference was that we have— next slide please, Ivan. So we add 22 rules, one by one, one by one, that I found from a real firewall while I testing. So, but even with 22 rules, it— the AI still made mistakes. Not as many, but enough. So what I did, I built a second AI that checks the first one. Stage number 1, the analyst. The analyst applies all 22 compliance rules, jurisdiction-specific, fact-bound, exactly like a junior lawyer that drafting. Stage number 2, a separate AI, 21 checks, catches overconfidence, false uncertainty, jurisdiction mixing. Exactly like a senior reviewing. Stage number 3, the finalizer. The finalizer merges, corrects, structures the output, ships to the user. Exactly like a partner signing off. So, who's made it reliable The— and accurate validation made it reliable. But what make it good? This is the UX itself. So when I had the prompts work and now I needed to productize it, I build it without engineers, also without UX design UX designers. At phase number 1, I just get the shape of it and I did the design exploration. I started with Claude, Claude AI, then I connected it to Claude Code, and then from that on I did kind of a ping pong between Claude Code and connected to Figma NCP and our design iteration between two of them, and then when it was look, uh, good for me, I connected to Laravel for, um, live prototype and deploy. I also connected between Laravel and Supabase between— for the authentication database and for the edge functions. Phase number 2, Cloud Design joined the party. I stopped using Figma, and I started to do kind of a ping-pong between Cloud Design and Cloud Code. And then when it was right for me, I connected it to Laravel for, again, for the deployment and build authentication. And then, uh, again with Supabase. As you can see, I did ping pong between the tools until it's right. The tools are off the shelf. The real work was the compliance methodology, not the code. So meet PapayaOne, the compliance AI we built so our customers don't use the wrong one. And let me show you, let me show it to you. Next slide. So this is the last part of the onboarding. I pick HR, my countries of interest, and get started. Two clicks, no forms, no question. The system knows who I am, and the old product is now built for me. Next slide. The other part that I want to show you, this is the homepage, not a blank check. Every workflow here at the predefined prompts also is for the HR role in Germany. This is the portfolio, my countries, my alerts. I didn't search, PapayaOne told me. Two clicks, two, two clicks at the start and the whole product changes. The ARM, the workflow, the alerts, every user sees a different Papaya one. Next slide. Check my contract. You remember that I talked about the 3 stages of pipeline? So as you can see, I upload contract, the Brazilian contract again, and then as you can see, it review it for compliance, thinking about like a junior agent, like a senior agent that reviews, and a partner that's signing off with all of these, uh, details. I have one concern about this, uh, contract that I just uploaded, and 13 are, uh, confirmed as compliance. And now I can easily take it to my legal department for further review. Next slide.

Next slide.

All right, so one more thing. Realichow works across multiple borders, so I asked Papaya One to compare Germany versus Brazil probation rules. Very quickly, the same 3 agents, same processes work for me and compare Brazil versus Germany and give me to decide what side by side and help me to decide what is the decision here and very easily I can decide and compare. So this is the work that they should do every day or another kind of role. That was what I built and the methodology that I used to build it. Now I want to turn it over to Sivan to tell you more about our go-to-market market approach.

Thank you, Khagit. Sorry, we have a little bit of a clicker situation here if you haven't noticed. So Khagit showed you what she built. And then the question was, like I asked in the beginning, when is it going to be ready to launch to our clients? We didn't launch it to everyone. We rarely do. We pick a small pool of clients, usually 5 to 10 clients, ones that we trust, the ones that feel like our partners, the ones that actually come to us often with compliance questions, because we know that they're gonna ask us the right questions and they're gonna give us the correct feedback and not just be polite with us. Framing every response is clearly marked as guidance, not legal advice. And here's what I wanna tell everyone in this room deploying a domain-specific agent. You don't launch it to all customers in one day. The product gets better from real questions, not from internal testing. And the hardest question isn't, is, does the AI agent work? It works. It can work that for one day. It can work after 4 weeks, just like Khagit showed you. The hardest question is, do I trust it enough to put my company's name on it? Do I trust it enough to launch it and tell all of my clients, you need to go to my knowledge base and not go to another AI agent in the middle of the night when you have that question come up in your mind? And that question takes much longer to answer than building the agent itself. We measure trust with 3 signals. One, are users coming back? Are they returning the next day or the next week with more questions? Are they asking harder questions? Higher-stake questions over time means trust is growing, but that's how we expand on it. If they come with higher-stake questions, then we know that we need to develop our answers further. And the last question, the last one is less outside counsel. Are they forwarding these messages directly to their in-house legal team? Because if they are, that means that they haven't built the trust with us entirely and we're not reducing the outside cost for them. When all three of these are in the right direction, that's when we expand. But most importantly, you also need to build a kill switch. If accuracy drops below a threshold in any country, we turn that country off until it's fixed. We actually pull the plug. We don't say, "Okay, we're gonna fix it as we go along," because that is what guardrails actually means in production, not just rules in a prompt, but a real mechanism that protects the outcome and your customers. So here's what I want you to take home with you today. The tools that Hagit used, Claude, Lovable, Supabase, you can sign up for all of them tomorrow. Most of the people in this room are already signed up for them. But the years of experience and 160 countries reviewing thousands of contracts, thousands of terminations, being dinosaurs in the industry, that can't be copied. And what takes longest is the people. Is building that trust and putting your company's name on it. Build the guardrails before you build the features. The AI tools will get better with time. The question is, can you teach it what you already know? All of this experience that you've gained, how do you extract that knowledge and put it into your AI, AI agent? In 4 weeks, we actually had a working AI agent. But it took us 4 months to build trust. The AI is the engine. The domain knowledge is your fuel. And you can copy the engine, but you can't copy the fuel. Thank you so much for your time today. We're actually out of time, so we don't have any time for questions. But if you want to come to our booth, please feel free. 211. Thank you.

Thank you.

Please give a warm welcome to the Chief Revenue Officer from Revo, Ali Ghotbi. Hello.

Mic check. There you go. Hello. Thank you for attending SaaStraw. I'm really excited to be here. My name is Ali Ghappi. I'm the Chief Revenue Officer at Revo. If you're not familiar with Revo, we're an AI-native revenue operating system, a revenue operating platform that's focused with agents that can help automate tasks autonomously, take actions, and proactively provide feedback and insights into your business. And I wanted to share a little bit about what we're doing at REVO as I scale out the organization, moving from sales activities to business outcomes with the help of agents. Traditional sales. I've been doing technology sales for 25 years. Has remained the same. It has been a myriad of activities, from getting a territory and focusing on where you should spend your time, to doing deep research on the accounts that you're going after, personalizing your reach-outs, then preparing for meetings. Then you run the meeting, you're taking notes, you're following up with meetings, you're putting it in CRM, and All of this is administrative. A lot of it's administrative. About 70 to 80% of time sellers are spending is on those administrative tasks. And there's a huge opportunity as you walk through the SaaStr conference, you're seeing everybody's putting AI into all the different activities. And that's something that is a huge opportunity as we scale out our sales organization. And so We found this as an opportunity to automate as much as we possibly can, and so we zeroed in specifically— clicker, progress. Clicker's not working. So we zeroed in specifically on— thank you— the opportunity management and what that actually looks like. And if you look at an opportunity and what a seller is actually doing on a day-to-day basis, first they have to prepare for a meeting, and that's doing research, going both onto the public web, looking for insights, news about the company, maybe reading their 10-K, looking through your CRM hygiene, your, your Salesforce, looking through all those activities of the opportunity past, what's happening with customer success. There's a lot that goes into preparing for a meeting. Then you actually have the meeting, and then you're taking notes. Then you have to take all of those notes and prepare them in a way that can both be actionable with an email that you're drafting back to the customer, customer, as well as putting it into the CRM, keeping that up to date, managing the business relationship, all the while getting coaching from your manager and trying to self-improve. We took a look at what are those things that we could automate the easiest and the most effectively. Right now, the, the opportunity of being in front of a customer and and managing that relationship is something that is key to a seller's activities. That's what we hire them for. That's what they're good at. It's not a coincidence that the best sellers are just absolutely horrible at CRM hygiene and doing all those administrative tasks. And so we spent our time focused on what can we actually automate on the administrative side of things. And so we looked at meeting prep. And we saw an opportunity to use agents to do deep research, both on the web, but also the entire signal of a customer's journey when they're interacting with Revo. So we put an agent on that. Then we took an agent and there's a lot of conversational tools out there pulling the intelligence out of that. So that was an easy one. There's— I don't think any doubt that the, that agents and AI can do that better than an actual human being at this point. The hygiene has been a common ongoing problem for years and years. And so this is something that agents can absolutely assist with. And then I can't be on every single call every day listening to my sellers. And so if we have an agent that's helping identify issues that sellers coming across day in day, whether it's objection handling or not filling out the right fields in MedPic or Medic, an agent can help. And we wanted to use an agent to do that. So that's what we focused on, and I'd love to show you a little bit about how our sellers are planning on working. So let's jump into the demo. So the first thing you will see here is our sellers are presenting represented with a single pane of glass, essentially one place to work, and they can look at their quota directly when they log in. How are they achieving against that? How is their team achieving against that? More interesting is that we've created these agents that have done a bunch of work that are either, uh, they've done it for you and they want to show you what they've done, or, uh, they are asking if they can go do some more work on your behalf. So the first one is prep meeting. So this is tied to Calendar. So for the day, they have an 8:00 AM call, a 9:15 call, and a 2:00 PM call. The agents have already gone ahead and done that deep research, both publicly available information or inside the overall platform. And they've come up with a summary and a meeting prep doc. So this meeting prep doc— Let me pull it out a little here. Will give you context on the company that you're meeting with, the key stakeholders, their title, their decision-making authority, their style and background, about the company itself, any recent news, potential pain hypothesis as a vendor selling to the customer, icebreaker questions that are personalized, discovery questions, objections and rebuttals, and meeting objectives. Now this document has been pulled, but is also refreshing constantly. So the agent is looking and seeing, have there been new signals? Whether there's new signals that are publicly available or recent emails or conversations that have been had. So it's a living, breathing document that is available to our sellers right before the call, available to everybody. The next agent is working on progressing deals. Agents looking at CRM and identifying where there might be areas to, uh, progress a deal. So for example, Amazon, 15 days, no motion. So there has been no movement in this account for 2 weeks. Uh, there's been no executive sponsor identified and is proposing a recovery email to be drafted. So the agent's done the work. And so if you, You hit Needs Review. What it's doing is it's searching through the contacts, searching through all the emails, searching through CRM and any sort of meetings that have happened, and it's gonna come up with an answer on why this deal needs to be progressed and how it needs to be progressed. So what's happening, it's identified again, 15 days in negotiation, no movement. The last movement was April 29th, 15 days ago, 3 follow-ups, no reply. It's also citing from the conversational intelligence that on April 29th there is a comment made that says, I need to run this by someone on the finance side, but no intro was made. That makes sense to me that we should probably get that intro made. It's already gone on behalf of that and crafted the email, highly personalized again with the context around asking for the introduction to the finance department. So the seller simply hits send and they are off to progressing their next deal based off of what the agent has suggested. The next agent that we went ahead and created was a CRM hygiene agent. This agent is going through while the sellers are sleeping and actually looking at all all the, again, emails, activity that's going on, all the conversations, and it's suggesting and proactively filling out the CRM to keep it up to date with the highest hygiene. And so things like next steps, opportunity amounts, executive sponsors, technology landscape. You can imagine how often this type of field does not get filled out, constantly as leaders badgering our sellers to go fill this out, this is doing it all the time. It's like getting a rebound when you take a shot and then getting it passed back to you so that you can actually continue with the game. It is, this is a game changer in terms of having high quality data to make decisions from. Also, what we've done is we've said, okay, let's progress deals, but we don't want them to progress win or lost without a seller's oversight. So the agent's also identifying deals that maybe need to be marked as closed lost, and so it'll pull all of those opportunities and it will ask you to disposition it, but it won't ask you to disposition it without giving you evidence and citing exactly why it thinks that you should make it a closed lost opportunity. And then it gives you the opportunity opportunity to approve or reject it. And so here it's identified 5 deals that need to be disqualified. This one is Twilio. It is, was in negotiation, but 90 days with zero engagement. Seller's just not even paying attention to this one. So it's identified that it's got, we've had 3 unanswered follow-up emails, and it's synthesized this all from email history. So I'm gonna go ahead and approve that. It goes simply to the next one. This one should be marked closed lost. It's proposing that based off of a sales, it was currently in the sales qualified stage. There's been no engagement. And one thing it's found here is that it's synthesized both email history and LinkedIn. The person that we were working with is no longer at the company anymore. And so we can make a decision to either progress and Reach out to the new person or maybe start over. So as the seller goes through and reads all of these and decides whether it should be moved on or closed loss, then you are automatically cleaning your pipeline. And this makes it a very simple way for sellers to really be presented with the information needed to make their decisions. The last one that I'll touch on is our Coach Me. So as I mentioned before, It's really difficult for managers and leaders to be in every conversation, but if we have an agent that's helping and understanding what we're looking for, we can use them to help coach our sellers. So we already identified there was a missed budget signal at Amazon. So that one is suggesting doing a recovery email that we just went through that example. There are examples of med pick areas that need that need to be filled out. It identifies where your gaps are and activities that you might need to do. And so this is what we're really excited about. It's been a big game changer. Going back into the presentation, there's two things that I wanted to just highlight, and then I'll wrap it up. Can we move it back to the presentation, please? All right, let's— all right, uh, before— can we go to before? All right, I'll move it back.

There we go.

So before, we were doing a lot of research. Now we have a living, breathing, dynamic document that is always up to date. Before, I was using OpenAI and Claude to send messages that I thought were highly personalized, but these all look the same. You've all probably received something like this. Now it's highly contextualized based off of all the conversations that we've had and the signal based off of what it's seeing across the platform. What that has enabled for us is I'm way behind on hiring right now, but we've already reached our number with half the reps because our sellers are 5 times more productive. They can handle more opportunities because they're having zero leakage. So from 10 to 15 opportunities, which is what I assumed any seller was gonna be able to handle at any given time without loss of, or degradation of the relationship, they're handling 50 to 75 at this point. So that's what we're really excited to share, and if you're interested to learn more, please stop by our booth. Thank you.

Hey! Hey! Hey! Hey!

Let's give a warm welcome to the founder and CEO of Fiscent Technologies, Adrian Murray, and the CEO of Launchpad, John Heune.

What a, what a wonderful week we've had so far. Thank you all for being here today. Here we are into the homestretch. So with all the talk about agentic apps and the shift from SaaS's, you know, seat-based and user-based pricing to the world of outcomes as a service where AI is actually doing the work for us. One of the big challenges that's facing companies that are selling into large-scale enterprises or regulated industries is how to get their AI-powered solutions from demo through procurement and security and into actual production with those meaningful businesses. So I'm John Hewin. I head up Launchpad for Pega Systems. We provide a platform on which companies, maybe like yours, are building agentic solutions that deliver outcomes, which may be an audit or a claim or even a closed, you know, sale might be the outcome. And they're building those on our AI-powered development engine, and then they're deploying those onto our enterprise-grade, secure, scalable architecture. And I've invited Adrian Murray, CEO of Fisent, to the stage with me today. To share his experiences building an AI-powered product and building a business around that, and that that business is now selling those applications into regulated banks and insurance companies and fintechs where they are being used at scale. So in the next 8 minutes, Adrian is going to show and tell you how he has built his Agentic Solutions and how he is having those deployed and adopted in those large-scale enterprises. Adrian, why don't you start by telling us what you built?

Thanks, John, and thanks for everybody for coming and joining us today. Um, as John mentioned, I'm Adrian Murray. I'm the founder and CEO of Fisent Technologies. We're an applied AI software company based out of Toronto. And for those of you guys that haven't heard of us, our platform BizAI is a platform that enables regulated enterprises to turn unstructured content into automated outcomes. Today what I'm going to tell you guys about is a concept that we and our partners at Launchpad call outcomes as a service. Really the whole idea of outcomes as a service is bringing together these series of technologies and the know-how to do something pretty simple, which is, which is get an outcome. And the idea there is we at Fisent bring bring the content intelligence layer, which we believe is a critical part of being able to address some of the more complex enterprise challenges and use cases that exist in the market. Launchpad brings the determinism and workflow and structured auditable platform that allows these things to actually run at scale. And together we've found and have proven and now scaled, uh, this solution to, uh, financial institutions across North America. So I'm going to start by taking that and kicking off a demo of an application we developed in partnership with Launchpad with BizAI at its core. This is an application called Reg ERI. It's, uh, focused on regulatory testing, so it's testing consumer complaints, and I think if it'll run. I don't know if it's going to run. Is it going to run? Hey, it's moving. You guys will start seeing this process. What I would highlight and want everyone to take out of this is you're going to see a workflow. You're going to see AI analyzing and processing this content as it goes through the workflow. But really, I would argue that's not the kind of only concept here. It's really the idea that these two things are working working together to get to this end state, which is validated, tested disputes that we can now, you know, reliably in an insured way prove to our customer, the financial institution, the fintech provider, that this dispute has actually been processed in a manner that's appropriate. And so really the idea here, I would say the big insight that we came across with our partners at Launchpad is that this is not an all or nothing proposition. I think there's a lot of folks out there that are sort of selling agentic AI as the answer to every kind of problem. Our argument is that the combination, John, right, of this sort of deterministic workflow with the probabilistic power of these models and the platform we've developed together is what actually delivers that outcome that gets a bank procurement person to actually buy and adopt and then run these solutions.

So from your perspective, Adrian, you're using deterministic logic for things that don't need to be creative and you're using GenAI in a more interpretive or creative fashion. So you minimize variability with workflow and deterministic logic, and you enable processing, ingestion, those types of things through generative AI.

That's right. And, you know, our argument, I think most folks that have been through this, if you have been through this procurement process with, with large banks or insurance companies, is that the procurement folks don't buy AI because it's interesting. We would love that. Them to do that, but they don't. They buy it because it can survive their process, which is information security reviews, architectural design assessments, and this whole litany of kind of tests and exams and compliance audits that you have to pass to actually be viable.

So you give the automation capability that comes with the Agentic workflow, but also then the confidence that there is a consistency and repeatability to the outcome that gets delivered. So we're not being creative when we're doing an audit, as an example.

Yeah, the way we like to put it is like our lesson, one of our lessons for folks is like, don't lead with the model. As much as we want to do that, lead with the outcome. And that's kind of what solutions like this let you sort of put together. In the case of, our customer, who is Kaufman Rawson, they're a, they're a well-renowned kind of regulatory testing advisory firm in the US. They do this work for some of the biggest fintech programs and banks in the United States. The lesson there is really that you can take kind of the messy evidence that they're dealing with in this process that you would've saw earlier in the workflow. You can combine that with the rules and, deterministic structure that something like Launchpad provides and bring in their SME knowledge and experience. When you put all that together, what you effectively get is a governed solution that they can actually run and deploy. But then I'd say like the up level to all that is, that's an interesting story and use case. And I think, I think John, you can kind of just cycle here.

Yeah.

But really this is a pattern that exists this across, we would say, sort of the entire financial services sector, maybe actually even much more broadly into other regulated industries where you can combine the power of these two technologies, the deterministic workflow and this probabilistic engine together. Uh, and we're seeing that firsthand. So we're solving these same problems in consumer complaints. We're doing stuff in claims, we're doing stuff in, uh, loan origination, really any type of workflow that has these two combination of factors, the sort of human inference that has to be applied, the deterministic logic and rules and SLAs and requirements, those two things together let you effectively build and deliver what we would call an outcome.

So I think, Adrian, one of the other things is like token maxing is a thing that people are talking about these days. And what you're doing with the deterministic logic And the workflow side is you're not burning tokens where you don't need to be burning tokens by, you know, driving generative AI. You're using the tokens in the places where you need to use tokens to ingest data, interpret data, you know, make things happen that pushes the process forward. But you're not burning them on, you know, the core running of the business workflow.

Yeah, I think that's, that's another great insight. And like everyone, I'm sure, is starting to hear a lot more about this. Enterprise budgets are getting kind of constrained and there's a lot of pressure on folks to like actually think a little bit more intelligently about how these technologies are being applied. As with all new technology, this is inevitably how things, how things evolve and develop. For us, kind of the, the proof is in the traction. So we've got a dozen customers we're doing this sort of thing with across 20+ use cases in production. We're processing month, hundreds of thousands of what we call virtual pages of content through this, through our service in combination with other technologies that are, that are providing the, the workflow foundation. Um, and we have a mandate from all of our customers to kind of expand and do more. So kind of how, how I'd wrap up for you guys, at least from my side in thinking about this is, you know, do when you're looking at a regulated enterprise and a, and a content complex business process, don't just think, okay, how do we add AI to this? Really focus on what outcome are we trying to deliver. And when you bring the technologies together to do that, it is possible to actually get this stuff to scale and, and in production. So John, I'll hand it back to you.

Well, thanks very much, Adrian. That was a very informative 8 minutes, so appreciate that. If you would like to connect with Adrian after the session, if you have any questions, you can find him, uh, and me, if you'd like to learn more about Launchpad, uh, we have a booth just across in the other hangar and would love to see you there. But also, if you're so inclined, you can get started building on Launchpad today. You know, just click on that, scan that QR code, it'll take you to our site, you can get started. We would be glad to connect with you, maybe introduce you to one of our, uh, forward deploy— forward deployed engineers and help you, uh, get going on creating outcomes in your business and growing and scaling your business with you. So thanks very much, everybody.

Please give a warm welcome from Scale Venture Partners, Jeremy Kaufman, from GCAI, Cecilia Zeniti, and from Inspirin, Alex Heinosh.

Good afternoon, everybody. Uh, we are excited to chat with everybody about the state of vertical AI in 2026. We want this presentation to be real. We've made a blood oath to the organizers. We are not gonna talk our book. We are not just gonna promote our companies. We're gonna talk about the real issues in vertical AI and talk about these businesses in the context of the types of issues everybody is facing today. So with that, we'll just introduce ourselves. I'm Jeremy Kaufman. I'm a partner at Scale Venture Partners. I've been there for 10 years, tend to focus on the application layer and verticals. We are Series A and Series B focused, and these are two of our entrepreneurs and love to let them introduce themselves.

I'm Cecilia Zanitti. I'm the co-founder and CEO of GCAI. It's a legal AI platform for companies and, um, huge fan of Jason's also because of the legal tech angle. I'm excited to be here. I am also, like I suspect many of you, uh, obsessed with kind of VC and software Twitter. And it's been fun to, to take that and apply it to the company. We have 1,600 customers and I checked this morning, a good 20% of our user base is actually SaaS companies. So I can speak to selling into SaaS companies as well. I was a general counsel for 10 years, lawyer for almost 20. Turns out legal AI is a thing, in case you didn't know. So that's been really fun and excited to chat with you all.

Hey everybody, my name is Alex. I'm the CEO of Inspiron, and what we're building is the intelligence layer for senior living. So we try and translate what's happening in the physical world, so what's the activities being performed by residents, patients, caregivers, and turning that into outcomes that matter to them, matter to them as human beings. So how do we extend the lifespan of people living in these facilities, and how do we make the operations run a little bit smoother. And that's who I am.

Awesome. So in terms of the agenda today, we want to talk about 5 key issues that I think are top of mind to most founders in the space and vertical AI today. First, we're just going to talk through trends in vertical AI funding. We're going to talk about competition and how teams can move faster. We're going to talk about the the, you know, age-old question of what it means to eat labor in the space. We're going to talk through competition between systems of record and systems of agents building on top, and then we're going to end with defensibility questions, which is very much top of mind today. So shout out to Euclid Ventures just for framing this. The data on this slide is from Euclid Ventures. They're a seed stage fund. They track all the deals done in vertical AI. So last year in 2025, there are about 2,500, uh, vertical AI rounds. So it's definitely an interesting space. So what do I take from this slide? It's a good reminder that, you know, legal AI is hot and getting a lot of attention, but half of vertical AI is actually healthcare and fintech. So it's just good to remind ourselves of that. And then on the right-hand side of the slide, you The numbers are a bit hard to read, but what I look at is I look at which of those verticals are most mature in their funding. So which ones of those are later stage? Legal AI and defense tech are where those big late stage rounds are happening, which makes sense. And it's a good reminder to, you know, early stage investors, let's go look at some of those verticals like agtech where maybe not as many late stage rounds have happened. So let's talk about what's happening at the application layer right now. You know, one of the interesting ideas at scale is we always say if you were a SaaS investor in 2001 and you just placed all your money in SaaS, you saw Marc Benioff and you said, yes, I wanted to invest in Salesforce. You made a lot of money. And if you said, you know what, I'm going to put my money into on-prem companies in 2002, you lost a lot of money. And what we're seeing in AI here is it's not as clear in all these core categories— customer support, legal. There's competition in 4 or 5 dimensions. There's old world companies like Salesforce adding agents on the left. There are these AI teenager companies that are started pre-ChatGPT but competing. Owen at Intercom and what he's doing in customer support is a great example of of that really making a lot of attention there. And I think the ecosystem underappreciates many of those companies pre-ChatGPT. We've got AI-native companies like Cecilia at GCAI, and then we've got these newer full-stack AI companies which are not selling software to accountants but creating an entirely new accounting firm. And the key question is speed matters. It's a competitive market. So Cecilia Yeah. When I think about what you did in the early days, I think you got off to a very fast start on two dimensions. First was that you sold to in-house and not law firms. And then second was that you took this product-led growth approach. So it wasn't top-down, down sales. So how did you decide to do that? And how does that contribute to the speed around the business today?

Yeah. So AI is a and specifically LLM-based AI is a once-in-a-generation tech shift, right? Or maybe twice if we have mobile phones and cloud. So whenever you have that, you can leapfrog the incumbents just by doing that thing. So we are incumbents in our space in legal tools was LexisNexis and Thomson Reuters, both of them over 100-year-old companies. So just the speed with which you can literally do anything is just not there. And so some of the feature gaps that I think from a software perspective would have been a problem just weren't because we could go straight to the AI. So from a speed standpoint, literally, I joke about this. I had a guy that early on that was a revenue coach and I can't understate how great it was to be thinking about that from day one. And he runs a blog, therevenuearchitect.com. But he basically— we were able to have users, beta users. And I still remember like clear as day this meeting with him. And he's like, we show him the usage stats and the engagement. And we had people doing 20 to— people doing 200 to 500 interactions with the platform a week in there for over an hour a day. And he looks at me and he's like, Cecilia, put on your big kid pants. The price is $3,000 a year. Go. And I was like, okay. And so that speed of like having the usage and the PLG first was like, okay, if you're using something for an hour a day and it is in the case of AI and lawyers saving you time, making your work better, we could literally just like see, okay, what was selling and then do the thing of like, okay, you don't really have PMF unless you're selling. And it turned out we did our first month, we did $100K in revenue.

Totally. Alex, I think you bring an interesting frame to this conversation, which is I think people in AI are talking about does the first mover always win or does the second or third mover in a market have a shot. And Alex is a good example of being the second mover, redefining the market, coming later. So talk about, you know, obviously we're talking speed matters, but you were the second or third mover in a category. You have more ARR and faster growth than anybody else in the category. How do we reconcile those two things?

Yeah, really interesting question. So I think for us, like, time was is actually an interesting factor here. We were like the second, third, maybe 15th mover in this space. Like, there were a lot of people trying to build here. So when I think about time, there's kind of like two dimensions that really matter. Basically, time for the market to mature, and then time for the companies to kind of narrow the aperture of what they're focused on. So all of the first movers that kind of entered our space, had this really, really narrow window that they were viewing the problem through. And, uh, you like architect your whole business differently if it's such a narrow window. There's just— you're doing everything for one particular workflow. So when we looked at time, we were like, wait a second, this is actually a much bigger opportunity, a much bigger problem. And it, it just allowed us to be a lot more ambitious with what we were doing. And then that subsequently resonated with the market. From like a market education standpoint, I guess the interesting thing here is it took a lot of education to let people be comfortable with like cameras coming into the bedroom, into the home. So the market had to learn that like this is okay, solutions are out there, develop budgets, you know, CEOs had to like want to prioritize it. So being a second mover over there was kind of like an amazing opportunity for us. And then if I think about like speed and like how we think about a culture of speed today, for us it's a little bit easier, right? I guess every single device that we get up on a wall, somebody's grandma lives for an extra 6 months. Like that's the basic mechanics of the business. So when you have that type of opportunity, community, it becomes like a moral imperative to go really, really fast. And that makes it easier for us as a leadership team to move away from, you know, engineering velocity and ship faster to like this cultural idea of like focus. And we're not going to do anything unless it's in service of getting these devices on the walls and helping people live longer. So it creates this permission structure culture for people to say no to 99 good ideas so that they can just do one great one.

Totally. Moving away, I guess, from the topic of speed, one of the key questions in vertical AI is how much bigger these companies can be than the prior generation of SaaS companies. In 2026, it's almost trite to talk in this audience about eating the work. You know, our own opinion at is we think that is a, you know, a massive market expansion opportunity. There's an inch— I, you know, worked at the Fed before venture. I'm an economist by background. And there's actually someone who did some work that showed technology advances usually priced at about 5 to 10% of the human labor costs. So I think it's a little bit of a stretch for, you know, the whole ecosystem to talk about eating the entirety of the labor. But I do credit to Chilia, you know, we're all talking about this in 2026, and this is a slide from your seed deck in 2022 or 2023 about the idea of vertical AI eating the labor. And specifically in your market, you had a very nuanced opinion around do you start by eating legal labor by focusing on company versus law firm? So maybe talk a little bit about this slide because you literally built this 3 or 4 years ago.

Yeah. So interesting stat about legal. When people talk about the $1.6 trillion market for legal services, who is it that pays for that, right? So your $4,000 an hour lawyer at O'Leatham Watkins, they're being paid for by a company, and ultimately that company is the one that has the legal risk, that needs the deal done, that has the thing needing cleaned up, that is doing the commercial contracts. You know, pick your legal thing. And an interesting trend that a lot of VCs didn't know, 'cause they weren't living it, is the growth of in-house lawyers. So companies have been insourcing that service for 25 years. So for the last 25 years, literally since before I graduated law school, companies have, in particular software companies in tech, Google came out with the term product counsel, and I was a product counsel at Amazon. Who here has ever worked worked with an in-house lawyer? Yeah, literally that's the entire room, right? And versus have you actually been deposed or in a courtroom, you know, sort of like, you know, like the, like, uh, whatever, Ally McBeal style? No, like that's not a thing, right? And so, I mean, maybe it is, but not for software companies in 2026.

But anyways, hopefully not for you guys.

Exactly, exactly. No, I mean, if you've been deposed, the best thing I can say is that it's kind of boring. It's not actually like that fun. So Anyways, this in-housing of the labor. So the idea was if you sell legal AI to law firms, it's kind of like selling Cursor or Replit to Infosys. Why would you do that? Like maybe Infosys will use it. That's great. But the actual source of the need, if someone at your company can use Replit directly to create the software, that was the insight that I had. And the numbers bear this out, right? So if you look at in the old days of like literally HP, you know, some miles down the road, Larry Sonsini, the founder of Wilson Sonsini, used to walk around there and it was like the big law firm. And they literally are called white shoe law firms. That's a thing. Like I didn't even know white shoes were a thing, but apparently they are. Anyways, white shoe law firms. What has changed since is like, okay, your product counts like Google, you're trying to launch, you know, Google Maps in a gajillion countries. There's privacy issues, there's whatever. You're going to hire that lawyer. So in an AI world, if AI can do some of the core threshold legal work, bring it to the company. So that was the insight. And now the fight that we're having or the debate that I think the market is having is what happens to that light blue box, right? So the old days was just law firms. The dark blue there is in-house counsel, which which today are 55% of legal spend by companies, is actually on salaries. And so then the question is like, in this world where agents do the work, who is gonna capture that value from the light blue box all the way to the left? My personal view, I mean, my company could, but no, just kidding. I mean, not kidding, kidding, not kidding. But anyways, you're gonna see some of that go to the foundation models, of course. A lot of it go to providers like me. You're going to see some of it go to providers of law firms. But that is like when you talk about speed, like tying those things together, going back to the like razor focus on the customer, capital C, like I come from Amazon, is like that's how to me that's my biggest advice for vertical AI founders is actually something that's like a consumer concept of customer obsession. So that's been huge for us is just like building building for the persona, building for the customer. And then as you're making decisions of where to put your dollar, you know, it could go to, go to us.

Totally.

I think one of the interesting things, you know, we've seen in the last couple of years in vertical AI is just the speed of adoption in a vertical and how you can go from almost low to zero usage, uh, one year to daily usage the next year. This year. The first person that I heard talk about this, I deeply credit Jake Heller at Casetext, another legal AI company, one of the first companies to kind of get acquired in this space. I remember he was describing how it's so interesting to build products in this environment because you ask your users one day what they think and then they change their minds literally 3 weeks later. And it's very much how much do you trust the user versus do you show the user. And I want to use GCAI as an example here because you guys saw very big changes in frequency of usage in your customer base. So maybe talk at a personal level, was that what you expected? And then how have you thought about this in terms of product development?

Yeah, so, so this chart is— I used to We give kind of teach classes on legal AI as part of our business, and we poll at the beginning, how often are you using some form of AI for real legal work? Not, you know, your ChatGPT for recipes or fantasy football, like actual, like real work, right? We asked this question 2 summers ago, and the answer was most people had literally never. And this curve to your question, like, is this expected? So the answer to that is a resounding yes. Yes. So, who has ever read the book Crossing the Chasm? All right, good. I love it. I love an educated audience. So, in that book, he talks about the early adopters being 2%. And look at 2 years ago, the people who were using AI hourly, it's exactly 2%. The next chunk in the sort of, um, the, the framework in Crossing the Chasm is 15% of people are sort of these early innovators. Look at the daily users 2 years ago, it was 15%. And so what has happened is essentially AI has resoundingly crossed the chasm where everyone now knows that AI and legal work are a good fit. Very similarly, by the way, to how AI and software development are a good fit, right? So 2 years ago, 2 and a half years ago, it was like, oh, there's a little bit of GitHub Copilot. And now, you know, I'm sure GitHub Copilot has users, but it's certainly it certainly doesn't have the imagination the way that, you know, a Codex or a Cursor or a Claude Code do. And so I think that reflection of like, can you spot where the puck is going and can you deliver for users in that moment? Like, that's the moment that we're in. And depending on your industry, you saw the VC trends. The VC trends I think are late to that. Like, look at, you know, Marc Andreessen talks about how, you know, if you wanna know what's happening happening in code or whatever, you gotta go, you gotta go talk to the 15-year-old hackers. Like, that's what we're seeing, um, in my industry for sure. And your industry, you know, probably has an equivalent where there's companies that are gonna be a little bit more at the forefront and companies that are gonna be less so.

Totally. One of the interesting topics that's come up across multiple verticals is the question of these newer startups on the agent side, whether you're in senior living, you're on the legal side, you're building agents and AI products. And then you also have these older world incumbent systems of record. You know, in the senior care world, in Alex's case, it was the electronic medical records. The folks on the right, you know, interestingly, in healthcare, the EMRs or the systems of record often specialize. So it's not one incumbent system of record, there's 3 or 3 or 4 that play in senior living, 3 or 4 that play in the hospital world. And Alex, you have an interesting strategic question as these incumbent system of records store the data, they've existed for 10 or 15 years in the space, yet at the same time, your camera is capturing new information that was never stored in that system. And your mission statement is alerting on falls, unsafe behavior, So you need to thread the needle on using your data as well as that data that exists in the incumbent system of record to kind of create this new magic experience. So talk a little bit about how you thread the needle, what data you work with, and how you actually work with those incumbent systems of record.

Yeah, a really good question. So for us, I guess we took a different approach than a lot of times what I'm seeing from like folks building and physical AI or vertical AI, we just don't think the systems of record should be messed with right now. Like, we're not trying to displace them or fight them or anything like that. The EMR today serves like a really valuable purpose in healthcare. Like, it's the clinical documentation system, it's billing, it's all of their daily workflows. There's just no reason for us to try and rip that out. But we know what it does well, which is like it's historical clinical documentation, and it tells you what somebody thinks happened. Some human being has to go and write into the system and remember what action was taken, and they write it down. Ours is just like completely different. It's like 24/7 ambient monitoring Cameras, radar, speakers, microphones pointed at human beings 24 hours a day, 7 days a week, building longitudinal profiles on them so that we can identify trends and deviations from the norm. That is just like not what the EMR is built for. And what we really thought was let the EMR be the system of record. We are increasingly becoming this system of action. We marry we carry those two things together, it actually works really well for our customers and they're happy. And I'll pause there because like there is nuance, right? There is like a moment here where it's like we want to be close, but I really don't want to be like a feature in somebody else's platform. So we have to do enough integration, we have to play nicely enough with our partners that our customers receive real value from it.

That matters.

But ultimately, you still have to own probably like one of the three things. Like, you need to own like deep workflow, deep data, or outcomes. And if you own one of those three, you'll probably be fine. But just don't fight them in the beginning. I just don't see a reason to do that.

Yeah, and it's funny, as, as you look across the hall behind us, you know, you got 50 or 75 software vendors there. And as a venture investor, I just feel like no matter what the topic is, sales tech, senior care, it's the same question in literally every market. I mean, do you start with your service cloud system of record, build an agent on top? Do you build the agent on top and then a year or two later build a system of record? And I feel like no matter what space you're in, founders are asking this question. VCs are asking this question and I actually, I like your answer around time. Like you might do something at time, time T, and then time T plus 1, you do something else. So appreciate that. You know, one of the questions that every founder is being asked today is around defensibility from foundation models. Like it is no secret that the rates of growth that we're seeing in the foundation models are unlike anything the ecosystem the ecosystem has ever seen before. It's stunning. It's stunning as a VC. It's stunning as an investor. It's stunning as a user of the technology. So internally at Scale, you know, when we talk about underwriting new deals, we're often asking ourselves, you know, how do we think about defensibility and what are the new types of deals we're looking to do? So this is a slide that we use internally. I mean, at a high level, we're saying defensibility comes in different areas, but also at the same time, it is an open question. And like, we're not afraid to admit that there are a lot of people on Twitter expressing certainty that I don't think the ecosystem has. Quite honestly, there is uncertainty here. So let's just say that out loud. So what we're talking about at SCaLE is companies like Alex and Inspiron, sensors in the real world, cameras, drones, robots. When there's a physical sensor there, We're seeing more and more businesses that are combining software and marketplaces. We funded a company called Paraform that basically gives free software away to recruiters so they can do their jobs better. In return, they join a marketplace. We're seeing more and more of those businesses. I think that's interesting. We're seeing— going into the enterprise, capturing context with forward-deployed engineers. You know, we VCs talk a lot about network effect businesses, and Cecilia pushed me on the prep call, which I appreciated, 2 or 3 days ago. She said, look, Jeremy, you have Forter and Socure, which are great identity-based businesses. They're AI companies, but they're not GenAI companies. And I thought that was a good comment, which is there are not too many GenAI companies where we're seeing these network effects just yet. We're seeing businesses like CodeRabbit and GCAI where the, you the moat is the user behavior, the context graph, the usage traces. And then we're seeing new businesses being formed that are saying, I'm going to build an accounting firm from the ground up. So we are seeing a lot of new ideas. So let's just say out loud that we don't know yet as an ecosystem. So interesting to hear from founders in the space what you're actually seeing. So maybe starting with Cecilia, you've been getting this question probably probably since day one of founding this company. So I'd love to hear, you know, how you thought about moats and what you did in the early days to stay ahead, and then what you're doing over the next 2 years to also stay ahead, because it's a question you've gotten, I'm sure, a million times.

Yeah, I mean, I think the— it's a really fun one to opine about, but I always come back to, okay, customer obsession and like what, What can you really build over time and what is more valuable to your customers? And so, I mean, the 7 powers, the 7, you know, 6 things on the slide, I have a hard time like kind of isolating them. But for us, razor focus on our persona and their needs. And, you know, over time, like we literally had a thread this morning and it was one of these, you know, as a founder, It obviously can be emotional, right? You have good days and bad. And we hit a revenue milestone that was very exciting. But this morning, a user just spontaneously started a thread that was like, I use Claude, but you'd have to pry GCAI from my cold, dead hands because it's just better. And okay, why is it better? It's better because every dang piece of feedback we read and we tweak the model, we tweak the product. We're constantly adding new things. And we have found that that razor focus on creating a great experience, creating, you know, details like, so for example, here's a detail. So we have contracts agents in Microsoft Word, okay, where literally we have the AI that can redline for you, it can make comments and so on. And I remember when we first developed the feature, maybe a year and a half ago, or developed the whole product. One of the engineers coming to me and saying, Cecilia, it's amazing. We figured out that when you strike through something, a whole provision, we can renumber the entire contract. And I was like, oh God, don't do that. Put reserved. That is what a real lawyer would do, is they would put reserved so you don't have to renumber everything. And something like that, You try Claude in Word for legal. It doesn't do that. And like, that's a little thing. But those of you who are developers can know, okay, each of these different AI toolings of like Codex or, or Cursor or any of these, you have preferences because they're better, right? And so if you think, okay, would you take a product that's 10% less good? I actually strongly disagree with the Twitter folks that are like, if your only advantage is that you're better, you're toast. I think that's false. Have you ever paid more for a better lawyer or a better doctor in your life? Of course you have. And that's a very natural decision that you make. And so for me, great product, customer obsession, and then doing it at scale. At this point, to your timing point, you know, we have 1,600 companies using the platform every single day. You know, if you want to start a legal company and compete with me, good luck. Have fun. Like, you know, like And so that's how I think about moats.

Totally. Alex, I know you're quite passionate about sensors, you know, given, given your company. And, you know, not only is the cost of sensor coming down, you know, a big motivation, I think, for what enabled your business, but it also points to the future and framing it in the context of moats. Maybe talk about, you know, why has the cost of your sensors come down so dramatically over time and what you think that's going to enable for the business.

Yeah, good question. I'd say there's like a few technical reasons why sensors have come down, in no particular order, right? But you've got the smartphone supply chain plus IoT just dramatically reducing core component cost For us, for companies that are trying to do this, that's a home run. Two, maybe even a couple of years ago, all inference was in the cloud, which made everything more expensive. You just have to do all the networking stuff that makes that actually possible. Today, that's not true. You can run these increasingly sophisticated models on the edge, which is kind of amazing. I guess 3, we got like as a world just way better at sensor fusion. So you don't need the most expensive sensor anymore. Like you can actually do all of that abstraction work at the software layer, which then means you just have all this additional capacity where you can take this. Like those are probably like the technical reasons that are interesting, but It's more of like the so what that's like actually kind of exciting. So like the chart here is like a— it's a gross margin chart, right? It's like, hey, things are getting like way better, but in service of what? It's just not in service of like financial outcomes and a gross margin slide that you get to show at a board meeting. I think when we think about it, it, it just changes the way that you deploy intelligence. So when intelligence was like really expensive and sensors were really expensive, you had this very narrow aperture, right? If you were going to put something into the world, you had to narrowly define the use case and you better be sure it's right because everything before that installation was really expensive. Now that's like not true anymore, right? So you can put these things everywhere into the wild and then all of a sudden, all of a sudden you start actually turning this flywheel that gets like really, really interesting and exciting, right? As the sensor cost comes down, I get more data. Data density increases the performance of the models. Better models are improving the ROI, which then gets you more customers and more devices on the walls. All of a sudden, like, it actually starts working. And it's like this convergence that is like what gets us excited to work, which is The cost of extracting value from the real world has totally fallen off a cliff, whereas the value that you can then return to customers with that is exploded, like up and to the right in a way that I've never seen before. That's a wildly exciting space to build into. And like, you know, 2, 3 years ago, everybody was telling us, like, don't talk about the hardware as a moat. Be a software company, it's like, no, it's way cooler now.

Oh, totally. Well, and to the timing point, I was actually product counsel on Alexa, and this opportunity of using sensors for seniors, Amazon saw it 10 years ago, but the AI wasn't good enough. So as soon as it is, you can get in there, and people, like to the moat point, I think actually being able to deploy the hardware in a way that you can get the software and AI gains, that's the moat.

Yeah.

Yeah.

And it's interesting as an investor, you know, the first time I met Alex, I didn't— I hadn't seen the prior generation of sensors, those super expensive ones on the left with the higher price point and a different form factor. So when I did the deal, you know, 20 other people in the community said this is not going to work. The sensor costs are too high. And it's like, it's so interesting in a market oftentimes times in hardware SaaS when suddenly it becomes doable because the sensor, the sensor has, has come down in cost. So maybe we've, we've got 3 minutes left and then I want to save 10 minutes for audience questions. So maybe we just end our conversation with Cecilia. You had a— you amused me earlier. You were talking, you gave me a quote, Ben Graham. This idea of the market is buy when depressed, Sell when euphoric. The idea that the market is a manic depressive. And sometimes in AI right now on Twitter with the rate of change, it must be quite— it's crazy being a founder, talking to VCs, talking to your customers. Maybe let's end by talking about how you navigate this and how you think about it.

Yeah. So the quote is Ben Graham, who was Charlie Munger and and, oh my God, Warren Buffett's mentor in 1949. And it was Mr. Market is a manic depressive. He's depressed sometimes, he's euphoric sometimes, and you kind of have to like figure out how to sort of sell within that. So AI Twitter, I'm literally telling you, like you'll see one line that's like OpenAI is dead, and then the next line is like, you know, software will never exist anymore. And then, you know, The very next tweet is like, Apple has lost its way. I'm like, what are you talking about? Like, literally, this is literally ridiculous to where it's almost like as a founder, you have to— for me, I come back to, again, the customer obsession, right? And this idea that, like, you know, if you've identified a real need that you're solving, do that. And then, of course, get the timing right. I mean, mind you, like here, I bought a house in Sunnyvale in 2012 and I sold it in 2017, did nothing to the house and doubled it. So like clearly, like you can't be stupid about timing. You should be smart. But the like allowing the like manicness or I don't know if that's PC, but anyway, allowing the like the vicissitudes of the market and in particular Twitter to rock you off your boat, like don't, don't do that. That's my main advice. And for us, like it's It's one of these things where I think it was Drew Houston or something from Dropbox said this, that part of your job as a founder is those highs and lows, is to figure out how to keep steady and one foot in front of the other. That's a little bit of the woo-woo California advice, but that's been helpful for me.

Alex, you're in New York, so what's the New York advice?

Exactly.

Same thing. I've got a 6-month-old kid at home that is like a grounding force for me, which is like very, very different. But I love everything that you just said around this idea of like maintaining a middle, that there are highs and lows. I also think it's, for me personally, just somewhat related to the fact that I have not slept the last 6 months, so like I just have this general normal even keel on like how to exist, and I'm rolling with it, and then eventually this baby's gonna start sleeping and I'll be fine.

Let's go.

That's where I'm at.

That's awesome. So on that note, we've got 10 minutes to take questions from the group. We were told that there's 2 mics in the center of the room, I believe, so we're happy to take questions. You know, I think it's great to hear from two real-world, two real-world founders wrestling with all this. So please feel free to raise your hand, step up to the mic. We are here to answer questions. Don't all rush at once.

I think this is working.

Yeah.

Um, have y'all had to justify pricing based off value or like world quantitative labor math that you're saving, and how did you prove it out to your customers? How did you adjust for it? Yeah, it's a great question because— so basically the question was around pricing, and it's everybody's favorite topic. I do want to pull the thread on what Jeremy said around traditionally the idea that software would capture about 10 to 15% of whatever value it places, I would say that's consistent with what we're seeing. Where, so for us, 14% reduction in outside counsel bills immediately doing nothing. Like you literally just start using GCAI and you're calling counsel less, or you are, when you call them, you have a much more pointed project. And so for us, like that displacement, the, the, the biggest tactical thing has been trying to go from the software budget to the budget around outside counsel for us or around headcount displacement, where we've got people on the platform where they had headcount for 2 additional lawyers and then they didn't need them because they're able to meet the needs of the company, you know, quicker by working faster and better. And so the hard part, though, if you're, if you happen to be in a fundraising process or thinking about it from like a business plan standpoint is that you're naturally gonna get the question, well, okay, would Claude or System X have displaced that price? But the thing is, is like, I come back to the point that I made, which is, okay, the, if the legal work is better, you're catching more risks, you're moving faster, you're opening more offices, you're taking less regulatory risk, whatever it is, like translating that to value, you kind of have to do it. But you also have to, again, that with that customer obsession, for us, I can't go in and say, you know, you're gonna white sheet your team and cleanly not have to hire any lawyers, 'cause then that's scary. So it's much more this like, okay, you can do more, and combining both the marketing and the pricing question. But on pricing, I mean, I think it's like, how do you price a grandma living 6 more months? I mean, my God, that's incredible. So he could charge anything, but I'm going to assume— I can't wait to hear your answer— but I'm going to assume that he's not like, all right, $10 million for your grandma. Like, that, that you can't market that way. So yeah, so how do you think about it?

That part would be, uh, $10 million would be amazing. $10 million entirely.

That's, that's our, our, our, our contract cost.

Yeah, I see. Yeah, I mean, the industry that we work with, uh, runs itself like multifamily property. So the mechan— mechanics and the math of it are actually like pre-done for us, which is really, really nice. Where it's like everybody cares about net operating income, everybody looks at their cap rate, everybody looks at the float, and then they say, wow, like Inspiron can add this much to my length of stay, this much to my occupancy rate. For that reason, I am willing to pay XYZ. Like it actually does normalize pretty evenly across the board depending on whether or not you're an ultra-luxury property or whether or not you're somewhere in the middle of the market. So we're, we're fortunate in the sense that like a lot of the hard work, like the math work, has actually just been, been done for us and we get to flow through it.

But see, again, to me that gets back to the customer obsession point. So his industry has metrics that they've met, and I would say using the language of your industry, super important. Where it's like for us, it's like we've thought about modules in the product where you get like your patent counsel or your employment counsel or whatever it is, where like you know the lingo of who you're selling to. And so use that to your advantage on pricing. The other thing that I will say is like when you're talking to sort of VCs, you probably have to talk about it a little bit differently than customers. You just do. And it's like they— and there is— he talked about decision traces. And things like that, there's like a— he actually gave me this advice, but there tends to be like a VC word of the week or something. Like just try to figure out whatever the word of the week is. Like, you know, like think about, answer the question of like how does this fit with the framework that I've mapped out or whatever my approach is. And like it doesn't mean the word of the week is wrong, by the way. I don't like literally— I think the decision trace framework and I I forget who wrote it, but it was a pretty good one. But, you know, it's kind of like use the frameworks how you would otherwise use them in running your business.

No, it's true. I find it so strange that sometimes we're all asking questions of the customers and then they're parroting back what the thought leader on Twitter said. And you're like, where is the actual signal in all this noise? So I think it's a good point. Yes. To the next question.

Hey, Rajan here. In one of the slides, you guys mentioned that you don't think that, at least for now, you want to get into the EMR or the system of record space.

That's also the application layer. So, two-part question.

One, do you think that that's an area that at some point you think is an area of opportunity for you?

And if so, what would the sign be where you think the timing is right for that? Yeah, definitely it's on our terrain map of places where we want to look. Like right now, we write a bunch of really, really valuable data into those systems of record that they then use to go and bill and charge and get paid. It doesn't make a lot of sense long-term for that to be a separate system of record if I'm the one generating all of the data that they then use to like solve that kind of like last mile problem. For me, there's definitely like an inflection point that we're noodling on when it comes to like, do we want to build into that sector? Our, our market is sub-10% penetrated for what what we do. So we are like very, very early days. And what we've collectively decided as like an organization is if we don't win the race to the wall, all the other cool ideas of stuff that we want to do doesn't matter. So we are myopically focused on getting the devices on the walls, motivating the team around this idea of like life extension. And then at some future state when there's enough penetration, then it makes sense for us to kind of like reconsider and say, is now the right time?

Thanks.

Sure. Thanks.

Hey guys, thanks for the insights.

I have two questions, one for Jeremy, another for the founders.

For Jeremy, how do you see like AI is making cheaper to make products, to accelerate and validate products? How do you see the future, whereas the actual value proposition of a VC, uh, where the founder doesn't need our money to, uh, to survive, maybe can do a lot of things with less money. And for the founders, AI, there's accelerating roadmap and etc., but is also accelerating your competitors. So how do you see retention and distribution challenges now with this, in this AI era?

Thank you.

Totally. So I love this question about how does the VC world change, because it's always a good question. And I remember I've been— I started in 2015, and I remember in 2017 people saying, well, with crypto we're not going to need the VCs. Um, and now with, you know, what AI can do, you know, do you need the VCs? You know, we actually are about to publish a really interesting study. Um, I actually really appreciate Alex and Cecilia contributing to this, which was we counted on the R&D side how much— what percentage in the last month of R&D was token resources versus people resources. And we're going to be graphing this over time because I think this is one of the big questions people are asking about building new companies and what it's going to look like. Interesting data point. Median company in the portfolio is using about 3% of their R&D spend on tokens. The most advanced are 15%. We think in the next couple of months that 3% is going to move to 20% and 30% across the portfolio. So yes, we— there's going to be less spend, you know, in some elements of R&D, but total spend is going to— is going to be there. And then practically speaking, you know, I look at the round sizes that are happening. Look, you know, being very direct, we are, you know, roughly a billion-dollar fund. You know, my greatest fear as a venture investor is not the venture industry and venture capitalists going away. It's just watching what's happened with industry concentration in the last year and a half. About 40 to 45% of the dollars in the industry have gone to really 6 or 7, you know, funds. We've seen this trend in management consulting and investment banking. And, you know, there's definitely an aggregation and conglomeration effects going on. But yes, I wouldn't— we're going to be using AI, but I wouldn't say I'm afraid of AI completely displacing the venture job as it exists today.

Yeah. So I think the question was around like retention and if everybody can build what you build, how do you, how do you keep a product? Well, I mean, we were chatting in the prep about taste and, you know, I'm a fan of the concept. But no, in seriousness, you can do anything. It's always been a thing that startups, particularly well-funded, you die of indigestion instead of starvation, where it's like there's too much to do. So actually the ability to focus and have that focus on the customer, on the concepts, on what's delivering value, to me, that's how we've done it. That being said, do I want to I love that we literally have salespeople in Cursor. We have like our, we have R&D attorneys, which is essentially legal engineers who are trained attorneys, lots of experience. They make code changes every day. And so that's allowing us to make our product better and to enduring advantage. So I would say everybody's got it, but then if everybody's got it, it's the same things that previously caused you to win, grit and speed and all these things.

No words.

Awesome. Well, thank you all for the questions. Thank you, Alex and Cecilia. Appreciate the time.

Thank you.

This is fun.