Ravin Thambapillia: From Palantir To Founding A Booming AI Startup - Credal.Ai | SVIC Podcast #23

welcome to S podcast talk about business Ai and comedy it's hosted by two former googlers uh my co-host is currently uh just uh uh going I don't know he's he's on a he's on a a road tour trip I need to I need to show I need to dive deep into that and show everyone how that all works and where he's going um but thank you for joining us we were just reacting to um Microsoft build conference and uh so that was be fun hey Ravine how you doing hey good thanks how are you hey good we are live Ravine is the CEO of credle AI uh a company at the Forefront of AI powered educational Technologies um and I think gp4 hallucinate because credle AI builds secure AI apps with your data workflow assistance Enterprise search and AI chat security security access controls and compliance built in so I will be giving uh gp4 a a thumbs down on on that introduction uh but before that uh Ravine worked at paler in as a deployment strategy strategist for seven years um he led in a team that worked on uh that worked on uh the built software that used to dispatch vaccines and Therapeutics hospitals during the covid-19 crisis and then he was the head of marketing at go cardless um and then he was the head of mobile at the hut group and then he was an industry Analyst at Google so former googler and he also has a uh ba economic and politics so he's a man of political science like I am so that is fantastic uh were you more of like for economics you more of a macro person or micro uh I was mostly in macro specifically in International Development economics actually but uh like dabbled a little bit in in the game theory side of micro as well oh fantastic that's that's awesome yeah speaking of Game Theory when the uh financial crisis happened and City groups stock was like going down to like nothing my dad was like you should put all your money City group right now and I'm like why he's like don't you understand Game Theory and and you're a political science major and how corrupt the system is they're going to get bailed out I'm like no I can't and he he nailed it and did very well so so that's awesome um C the analytics says I pay for just two things in the web chat GPT and the sik channel for its videos and Associate Discord group it's a unique form for those interested in the advanced aspects of AI technology and business Jordan hunt says it feels like an amazing community of people who are stupidly passionate about tech and are just straight up nice and absolute true of information and life experiences plus it's a great Hub to keep up with the chaos that is AI at the moment especially when life is keeping me busy my anti-ad internet Theory home we talk about robotics in here we have a channel for AI place for research people share their research papers in there memes and comedy gaming entertainment investing documentaries name it we got it hit that join button right now or support us on patreon so you can get access to this community yeah so we here to talk about your your company uh credle uh is it credle or credle crle actually yeah we say credle crle Okay fantastic so why don't you give us just a quick overview of it uh and I'll I'll do a quick just screen share of the website and everything so people can get an idea of what to expect so you take it away so as it kind of says on the website like crto helps uh Enterprises build secure AI apps with their data so we come with a bunch of different data connectors you can kind of see them on the screen actually that help uh Enterprise connect different data sources uh as not just the content but also the permissions about who at the Enterprise can actually see each of these different data sources they can then Define kind of security policies on top of that on not just the permissions but also like does this data contain pii does it contain Phi does it contain information I might not want to use with a large language model right um and so when an employee is asking a question like let's take a really simple example might be um what's the company vacation policy right uh cre can automatically detect the relevant information across the Enterprise to answer that question bring it in as context and then um sort of feed it to an LM to get a response but redacting any of the sensitive pii Phi financial information before the llm gets it and then unreacting it on the way back so the users sort of respones is um uh you know completely unaffected more complicated example might involve like building so some of our fintech customers build their entire like kyc and KYB process on top of crle so there what's kyc KY I have a I have a first grade education great thank you for calling me out on that so kyc is know your customer KYB is know your business so if you're a fintech company you serve either end users or or businesses you have to be able to prove to Regulators that hey this is not a terrorist this is not like a drug C tell these are like legitimate companies that have like a legitimate need to use you know our bank or whatever else it is um and so when you onboard them you need to be able to do a whole bunch of research to say this is all the research I've done this is all the data I've collected about this customer and this is how I know it's like definitely not a drug cartel right and so run those processes through credle bring in all of the different information that they have about the customer bring in all of the different information they have about the policies like a drug cartel might be like a really obvious example where you obviously don't want to serve it but there might be a slightly more complex example where it's like this person is the nephew of the Prime Minister of you know uh Egypt or whatever and there it's like you know it's not necessarily a bad person but like there is some risk involved in supporting that customer and so you know how do you make sure that your company's policies are deeply understood by the AI as well as like the specific specific risks associated with this specific customer and so bringing all of that together uh in one place such that an llm can really deeply understand your business but you don't have to hand over all of your Trade Secrets or all of your most sensitive data to these like kind of scary third party models that might deep fake deep fake your voice and monetize it so deep fake your voice Scarlet Johansson's voice huh yeah make a little bit of a joke about SC scarl Johansson's voice but yeah yeah what we uh this is super interesting because um I was I was thinking about like because the provisioning part of making sure people get the right information at the right time right um and approving that and whatnot I mean I think that's super valuable because I think we've all I mean you're imp paler I was at google Jonathan was over at slick deals and qu there's been situations where he probably needed data and couldn't get it because someone's out in the weekend or something or it's a weekday and you can't to file a ticket and for your assistant to be able to do that automatically for people or redact information that's that's really amazing um what what have your customers been saying how first of all one how did you think of doing this first because I know you're in paler for seven years so security is like your top concern so maybe you can go into what inspired you to create this yeah I mean it it's funny like the the story of our of our businesses uh is kind of an odd one Jack and I were both at palente for like over five years a really long time there in my case seven in his case five five and so we were working on really hairy data environments like Jack was at the Department of Defense working on you know computer vision models in uh combat and I was working at the time just before we started credle on our work with HHS for the pandemic response trying to stop America's hospitals from overflowing right and um in both of those cases as you can imagine you're dealing with like insanely sensitive data but this was also the time when the sort of generative model like wave was just becoming it was becoming clear that this would be extremely important so in 2021 I think it was gpd2 was released and you could only really play with this if you were like a software engineer right you had to download the models you had to find a way to like actually run them on your infrastructure it was a huge pain but if you could do that you were exposed to how insanely powerful and how insanely cool these tools were and so Jack and I were both um in these like super strange scenarios where we're trying to deploy these generative models in the most insanely high risk averse like um environments you can imagine and so we both we had dinner at this Vietnamese place in Chelsea in New York and we were just talking about how on the one hand the models are insanely powerful and on the other hand you know it was so difficult to persuade the Department of Defense right to just actually deploy them um and we realized like look we're exposed to this problem a few years before anybody else is but like oh boy every single Enterprise in the world is going to need to figure out these security and governance problems um so I think you know started from that Insight which was very much the exactly right Insight but actually I think we derived the exact wrong conclusion which is like okay let's go build an AI application and let's bake in some really cool security features like help sell that um and so we went off and built like kind of a bad idea right we built like an AI Chief of Staff but we built it with like all these very sophisticated data governance permissions mirroring um pii controls all that stuff and so when we went to sell that in the market what we would see is people would be like well the product is kind of interesting but I'm not that interested in it to be honest but the security controls are amazing and I really need to implement this for these like three tools that I'm work on in House at the moment and can you teach me how you guys did that and so that's when we realized like oh wait we should just scrap like all of the stuff that isn't the security and data stuff that we've done really well and we should just sell like the secure data platform as a service and and that's how the business got going that's amazing so how many customers did it take for you or customer conversations before you decide hey you know what we need to change what we're doing here yeah good question I don't know how many is the right question but more like there was one very specific one where it was like really clear and what happened was we had this customer latchel they're like a property management firm and they were using the old product right and they were paying us for it but like not very much and we were trying to get them to pay us like a lot more for it and uh he recorded his conversation with his executive team who were all using the product and in theory all getting value out of it and he was like right now this is coming out of like my sort of innovation budget you guys need to like take over the the contract and he went around the room and he was like how much you going to pay for it and for context like the product is working at this organization it's doing its job like pretty well the customers are actually generally like pretty happy with the fact that it's useful and he goes around he's like how much is it is it worth to you and each person in the room is like well maybe I would pay like 20 bucks a month and the next one's like yeah maybe I would pay 10 bucks a month and that's when I saw that conversation I was like oh damn like we're off by like you know two orders of magnitude here you know um and I think when you realize okay we've actually built a product that people like and they it does the job that it supposed to do like pretty well but it's not valuable actually right it's not solving a problem that they like deeply care about and that's when we realized like okay we need to change what we're doing and we looked back on all of the previous customer calls that we had and we like reflected on the feedback that we' got and we were like oh yeah like everybody keeps asking about how do we build all these amazing security features and then like way fewer people are actually asking about like the functionality of the actual product and so that's kind of how we realized that we needed to change direction and how we settled on what we did nice move now um did you were you still boot were you bootstrapping at that point yeah so at that point we were um I think think we' actually gotten into YC actually so we had a little bit of of money from YC I think they give you like 125k but we hadn't spent any of it at that point so we were um we were basically you know just running on our own savings from palent wow yeah I like the financial discipline because it allowed you to say okay this is not working let's let's move let's let's let's rebuild so how are you able to get this V2 out how long did it take you to do that yeah we we basically sort of sold the V2 long before we built it so the the irony of that that pivot moment is that it came when I had 150 investor calls booked the following week right so I had one week um before I was going to start taking these investor meetings and the reason for that was that we had YC's demo day and like the sort of tradition with y's demo day is you try and raise your seed round before um you do the demo day and I like you know suddenly had a week and I was like there's we have no business right we're scrapping the old product how the hell am I going to take these calls and like um uh try and you know make it useful so I pushed all the calls out by a couple weeks and um instead we just tried to sell the product we like let's not even try and build it like selling security tooling is like a relatively slow process right you you it's hard to land a 100K contract in like a week um so let's just make as much Pro progress as we can on the sales br and then if it looks like we're getting traction on the sales front then we can like um rip out all the crap that we don't need uh from the product and and sort of productize it and that works well I we're going to turn into a YouTube short I love this because it's very easy to say I mean because you're it's easy sometimes like for business you dive into what you like doing for the business so if you're marketing or sales you go to Marketing sales for engineering I'm just going to code and not talk to a not talk to a customer but for the business to thrive to lean into things you don't like doing and so it you could have said during your pivot let's just go build this whole entire thing out and then later go to the market and you're like no let's first just sell and see people like the concept and then we'll build along as we go that that's that's genius yeah and like honestly this is a big learning for me this was not a skill I was inherently good at like really the fact that I had like two weeks to like generate some revenue for this product that didn't exist like really forced me to just just accept reality as it was I couldn't Kid myself into being like Oh let's just like go cod the product no one will buy it for 4 weeks or 8 weeks or whatever but it doesn't matter like I had to make a sale I had I had to bring customers on and um that like needed discipline forced us to do the right thing actually which was don't bother building anything go out there speak to customers figure out what they need to buy and then like you know figure out how to adapt the product into that and that was exactly the right thing to do at that stage in our business fantastic now uh when you you finally got a customer was like I like this when can I get it were you're like [Music] ah so yeah it was scary right so um and there were a whole bunch of things that were scary about that so we um we sent out a bunch of like cold outbound email like that day when we were like okay we need to do security uh we had two folks respond um to that outbound one of them was like oh we will we're actually literally just about to try and build this product ourselves in house can we jump on a call so we jumped on a call with them the next day the security Guy brought like the lawyer the finance guy he brought like the whole Kaboodle from their team and this was like a big you know 2,000 person organization uh market cap of about you know 5 to 10 billion doar it's a really serious company they're bringing in like the big dogs to this like random two person startup so that's when we knew like okay these guys are actually serious that they do really want something in this space um and that gave us a little bit of the impetus to be like okay like let's really try here so we got them to sign an Loi with like I think a $60,000 price tag attached to it um and then um alongside that Loi we had like a statement of work and that had like the four or five features and they were just like honestly really good about it they were kind of like okay we need this feature by this date and this feature by that day at that feature by that day and honestly you know it was kind of good we could surprise them in a positive way even though we had almost none of the actual product built when they signed the LOI like they were still surprised at the rate of which we could ship things and you know when you need to get your heads down and ship you need to get your heads down and really ship but um you know luckily you know we were in a position where we were able to do that and we were working with like very understanding Partners as our like very first Enterprise customer fantastic now um two questions one uh it was still a two-person team when you closed this closest deal or yeah it was still two person team literally just me and Jack okay and so and then um I had a question in the comment section how many customer interviews did Rin do to have a good signal on what to do yeah it's so funny like the before we started the comp at all right we did 40 like actually somewhere between 40 and 50 customer interviews we spoke to people and those actually all turned out to be pointless and a waste of time like we got exactly the wrong signal from those customer interviews that like yeah I'm managing you know projects is really hard and like productivity software sucks and all of that stuff was true but what we didn't get was the important signal which is that like finding budget for productivity software is basically impossible and so you're like you're very unlikely to do it as a startup unless you solve like a very very very specific paino um and so we didn't get that signal because we were speaking to the end users not the like actual buyers of the software um and so we built the wrong product and then we were trying to sell it right and then in the process of trying to sell it that's when we actually learn stuff right and I think one of the big things that's important to remember is like when you're doing a customer interview there's a big difference between finding something that user says is annoying or a user finds painful and something that will actually motivate a 1,000 person Enterprise to buy software from you right um and so on that on the sales side we must have spoken to like maybe 10 15 buyers but we very quickly heard like the very same pain points around I'm terrified about you know sending my data to open AI I really can't let you know the permissions of the underlying Source systems get ignored when I build my internal llm story and those those pain points are like Universal you don't have to have that many conversations the key is like make sure that the pain points that you're finding are actually like hair on fire problems they're not just like ah this is kind of a you know this is a pain point I don't like it but at the end of the day I'm not going to put my neck on the line to force my organization to buy it from some weird startup that they've never heard of you know Robin I'm a little curious about your experience just kind of working with it departments and uh office of CIO my own experience is that uh chat GPT kind of like did this weird thing especially in the midmarket where all of a sudden these it departments that had been mostly kind of like ignored in the area of SAS right like they got a really long day off for the last six or or 10 years they were kind of relegated to M managing OCTA for human resources they didn't really have to make software work very hard on integration because the the SAS Enders were doing such a good job of kind of managing it themselves uh but but it seems like AI kind of threw a wrench in there where all of a sudden you know every coo ask their CIO hey how much of this stuff can we make now like how valuable is our data how can we start you know bypassing our traditional uh business intelligence system and then and then running our own queries against our own data and so it seems like more kind of traditional infrastructure enablement software like yours is going to be really important but but but my take is a lot of it departments are understaffed they don't have the talent necessarily to kind of Drive Leadership for these decisions right but you've now had this perspective of talking to them about AI for the last three years so you probably have a much better perspective than I do yeah I think it's one of those things like different organizations have approached it differently I think historically the technology organizations have always believed especially like the sort of cloud native technology organizations have always believed that like having your data in a good place that you know it's easily easily accessible and that people can make datadriven decisions really quickly like those organizations are really well set up to like benefit and adopt AI um really quickly because the data isn't a place where you know they can actually make those decisions quickly it's the more traditional organizations that have not um been able to get ahead of the sort of cloud native question of like how do I get my data in one place those organizations are finding it much harder and it's much more of a struggle and it's much more of a battle and that's also where I think it has been most underresourced right um for for a long time and so yeah I think one of the things that's like really really interesting about this change is it it is quite similar in some ways to the shift to Cloud which is that you know the shift to Cloud took like the data exhaust that your business systems produce used from like you know just this random thing that you know existed to oh wait this actually might be the most valuable thing that I business has to like today where it's like this is definitely the most valuable thing that our business has right because the software can be replicated probably really easily in the next five years by a team that's like half as big as the team that we had um but they're like powered by AI or they're like um you know benefiting from all the existing um you know inov that's happened uh in the open source market as well uh and so the real sort of value of that data that you have that's your Mo right that's the thing that makes you valuable it's not so much the fact that you spent five years or whatever building software before it's that you collected this insanely valuable data asset in that time and now yeah the cios the cdos which is like a a term that didn't really exist that much like certainly 10 years ago even five years ago right this cheapap data off a guy or girl obviously like that character is now suddenly extremely important um and I think you know the shift had already started to happen in the cloud native world where people organizations were starting to realize my data is important but this is just injected like you know a huge shot of adrenaline into that move where it's like okay our business in order to survive needs to take its data seriously needs to think of its data as the core mode that creates the business value yep that totally makes sense great answer you um you mentioned Finance I would have to imagine Healthcare is also right up near the top and then uh how much traction uh to Total direct question but how much traction have you had with Enterprise marketing uh when you say Enterprise marketing do you mean us selling into marketing teams or do you mean Enterprise that's right yeah like the CIO is a major marketing departments who also have their own data sources but have all the complexity that you just mentioned right that didn't exist 10 years ago you could do whatever you want you could send emails to whever you want now you have to depending on what Market you're in you have to be extremely careful about how you mix the data and what you do with it yeah so actually it's a great question and um you sort of hinted at this in in the question but the market is a really big deal here so if you're in Europe and you have gdpr related marketing concerns or even these days like in the US where you you have to worry about the CCPA thinking about even something as simple as like a user's email address which may not seem as scary as you know whether the like user has like a healthcare um concern or not but a simple thing like an email address is now considered regulated data right and you have to be extremely thoughtful and extremely careful about where that email address is going who's getting exposed to it um even internally at your organization like who can see the email addresses names uh addresses of uh your end users even your employees right we have some customers where the employee names are sensitive internal to the organization right so I as Raven cannot know that I have a colleague named Jonathan BR right you're probably bored looking for a hobby or maybe you're just trying to escape the dread of a job you don't enjoy whatever the reason thans had you covered than is a 3D model Community for designers engineers and enthusiasts it's the world's largest 3D model search engine databased boasting over 20 million free and premium models for you to search store and collaborate on they're offering you a fantastic 15% discount on your subscription for the first 3 months which is absolutely fantastic there are three stores you can choose from to use this discount for Terry Divergent they offer these awesome little Planters and faces and watering buckets that you can print out put in your desk and look really bougie and cool I'll probably print out one of these but I'm not very cool but we'll see what it can do or there's big bricks where you can print out all these awesome Lego figurines you can have them for yourself or you can share them with your kids or your nephews your nieces you don't want to give them out to random kids cuz that's going to be like kind of weird you look kind of a creeper uh and then there's Pokemon who doesn't like Pokemon you can get Pikachu Squirtle and all these other cute cool little oh that's like a Bubba Fett mask on an Eevee that is fantastic anyways headover things.com either go to super crazy print store or go to their store on big bricks or the Tera Veron store and use promo code svic and get a 15% discount on your first 3 months hope you enjoy talk to you later bye and that's at the extreme end of the sensitivity but um it is like you know it is an indicator of how seemingly simple data can become really sensitive really fast in regulated environments and obviously marketing data is a big one for that because you're holding customer names customer addresses customer email addresses birthdays all that kind of stuff Jordan I can see you've got a question but I just need to throw in no please continue back in my day you used to be able to do anything you want with somebody's social security number there was no regulation it was was very was very UNC consumer friendly but now we've swung very appropriately and very progressively towards a lot more consumer protection and so the the challenges that all these marketing departments have is okay you gave me your email and you gave me your address but I have to keep those in two separate places and I can never conjoin them right and I can't even pretend or act like I know what Pro what product you prefer depending on the market that you're in I'm not allowed to keep personalization data or if I do I have to separate it wrap layers of around it and so yeah I'm I'm just replaying all the conversations that I've had with all of these people who are facing that level of complexity and again marketing departments sometimes have their own CIO and IT staff but not always uh and this this ends up being a very very you know the the internet brought with it and and all of this great technology brought with it the the uh the expectation that we would have really really good personalization but with personalization becomes you know also brings privacy breakdown exactly we've we've seen where that goes if left unchecked so it's interesting that you're now bringing this potential solution to these people and say hey I can fix all that for you you can still send really good emails you just have to respect the data where it sits exactly yeah I think one of the other things that's actually really like challenging for all these marketing teams is a lot of these regulations require like right of deletion right so uh end user can come to you and say hey get rid of all of my data now if you collected data thanks to the EU right thanks to the EU you collected the data in HubSpot and then you eted it into your Snowflake and from Snowflake it went to you know your mongod DB and then it went to your Google Drive and then it went to you know like these 27 different Source systems and you have no idea really like across all the different parts of the organization where it might have ended up Suddenly someone comes to you and is they're like can you delete the data that I sent to you in a HubSpot and you're like well I can delete it in a hopspot but there's like 27 different systems and I have no idea which of those it may may not be in right and so keeping a catalog of where that information is um such that you can sort of deal with these sorts of regulatory requests is really challenging for those um those uh marketing teams especially in um you know especially in the AI world where you've added these 20 new Vector databases and 14 new like AI applications that are all trying to access that data this is awesome um Robin where would you like to go next year on this interview because I want to like what are things that you haven't talked about yet in other interviews that you're really excited about where are things you're seeing going on the market and you're like ah I wish people would not do this anymore just wherever you want to go yeah totally so I think one of the things that's been really interesting this year like honestly not even this year last like two months three months has been but we finally started to see the emergence of the first very serious like agents that actually work um and I think the honestly it was kind of a surprise to me I was a little bit of a like you know I obviously knew AI tools were amazing and like started an AI company because I was so excited about them in 2022 before uh chat GPT was even a thing but you know I was still pretty skeptical that we could get you know really sophisticated products um to market doing extremely sophisticated things now it's still early candidly but we've seen for example uh kyc processes uh kind of like we were saying earlier where a regulated entity like a large financial institution has to make a regulated decision about whether or not to allow this customer to open a bank account right and today like we have a big public customer that is making 92% of those decisions through an entirely LM agent um approach right and that is a really really mindboggling development and I think like the part of it that's really interesting to me or the the thing that actually is perhaps most surprising is what drove the ability to go from these like toy use cases to the like really sophisticated this is driving the core operations of this business and like radically reshaping like nature of what this company does was not actually so much that the model got a lot better because between the adoption of the first version of gd4 and the latest the you know the benchmark scores improved a little bit they moved a little bit here Claude got better for a bit they got better for a bit but you know it was actually like not a significant Improvement in those score benchmarks what actually mattered was all of the work in bringing the data in cleaning it up getting it in a place that was led ible to the AI and so what what I mean by legible to the AI is like all of your data today is stored in some system which is typically meant to be read by like a human being it has like a bunch of weird formatting rules that like AIS don't necessarily understand um you know it's got a bunch of you know there's a bunch of Noise We every CIO in the world knows that like the data in their Enterprise is invariably afflicted by like all of these difficult quality issues and so actually the investment that uh that these companies need turns out to be how do you take that data and how do you make it legible and clear and comprehensible to a large language model and that's actually more important than uh spending 100 million doar on like a big training run for your like super Advanced sophisticated in-house built uh large language model which you actually don't need it turns out what you actually need is just like really high quality data and I think that's the thing that I wish the market like would wise up to faster right is stop trying to build your own in-house model stop trying to you know run these like insane fine-tuning jobs on low quality data um actually just focus on get getting your data in place um and the market you know the ocean of the quality of the large language models is going up that will continue to go up with or without you basically the thing that won't go up without your investment is the quality of your data um and that's the thing that I think I've just begun to see like oh wait in order to unlock these agents it's not about gbt 5 actually it's about really really clear high quality uh data a amen so well said because I remember when I first got access to chat GPT and then I heard about fine-tuning I immediately thought oh I need to fine-tune this cuz my data is so special great and then it was like no actually first work on you're prompt oh prompt worked out well I'm in good shape or first actually make it so chat GPD can easily get your information and clean it for crap because a lot of databases are garbage and I I mean Jonathan and Ravine you probably can testify to this you probably get access to Enterprise customers and it's just a complete disaster internally and they expect that you're going to be able to like fix everything for you it's like no you need to there's some things that you need to take care of yourself here you can't just depend upon the model to fix for you right so EXA I think one of super interesting is just slack as a data source I think slack is has been really cool for Enterprises because especially technology Enterprises because it has just been a place where a lot of information does just kind of crop up there right but the problem is a lot of the information in slack is either untrue or it's out of date or it's stale and it can be super misleading for a large language model right so especially for like folks who are hoping to build their entire AI strategy around like okay like let's just give it access to all of the data and hope that it can like do a good job actually you have to worry quite a lot about well you know is it actually going to accidentally read untrue information or or stale information or whatever and um just provide like employees with completely incorrect uh directions or customers with you know fake discounts or that kind kind of nonsense right it's so true because um I was launching this move work spot at Salesforce and we had just um oh God tens of thousands of HR articles all in various states of being updated or not and of course A lot of them weren't because people leave and stuff happens and one question I we get all the time is like well if you turn this thing on what if it starts serving up bad answers and I and I would i' I'd say it this nicely but what I wanted to say is like well you got to get up get off your lazy ass and clean that damn data cuz you're going to see what article it's served if that was bad then you need to fix it and close the loop and so I think this has been Central in the AI agent aspect is no we're not at the point of just tell an AI agent to go uh cure cancer and come back and we're good right we're at the point of hey AI agent can you go do the groundwork for something for me and then come back to me before a decision is made so I can make sure you're in the same track or clean up what you're doing and then let's keep the process going like we're like we're co- programming or we're co-piloting a plane or something so yeah 100% I was just going to say I think what you guys what you guys are talking about though really kind of infers a knowledge and an expertise in AI that in my impression seems to be mostly still lacking like at at the you know at the bare minimum I would say there's an infer of knowledge where like there's some questions that you should really ask you can only ask of your own data but most other questions were probably better answer answer by something that's been trained on much larger data sets I I talk to a lot of people who run very large communities and so this question is very Central to moderation there's uh a lot of things that you can only ask your own data like what are the people in my community thinking about this what are they thinking about that but if you really want to ask those questions you have to have a model that was trained on Far larger data sets so that it can recognize things like emotional response and basic human language pattern recognition and you're never going to have that much data to train any anywhere near accurately on those two vectors and so it's but it's a real struggle to get people to understand that there are different kinds of questions in a you know a business intelligence system querying your own data small data set structured queries is entirely different than building an AI model to query your own data set which which you really need an AI model built somewhere else on much broader data but it's been a struggle for me to get people to understand the difference it really seems like your company is going to end up being an enablement to people feeling more comfortable trusting their data outside models where they're going to get a lot more value yeah I think like I mean to me it's really obvious right would you rather hire like an incredibly smart person who uh has access who can you know read the books that you need them to know to acquire the knowledge would you rather hire someone who's only ever read the like four textbooks about the specific topic that you that you have and has never been able to read anything else now there might actually be some like very very very specific use cases where like you actually want the person who's only ever read the four textbooks but that is going to be like .1% right of your AI employees and the 99.9% of your AI employees are like you want someone who's smart enough um and who's like got access to the materials that they need to be able to do the specialized task um rather than you want someone who like Knows A specialized task but is actually like not capable of reasoning at all about how to do it right you you you made me think of something interesting I have to imagine I'm just uh I'm putting my old Enterprise sales hat on I I have to imagine in your sales cycle there's a little bit of Education just kind of about AI still uh depending on the customer and is I think you kind of answered this partially earlier when you said there's you know companies that are very well positioned in this they're well past the decision of on premer cloud and they've kind of they realize the value of their data they're kind of getting this early are you are you working at all with customers that aren't quite there and you're having to like be the education source for how they even think about how to use AI in their data yeah so there's actually really interesting pattern here which we've seen which is that there is a type of customer which has like very very smart employees and um but historically has not had much data right because the the data that they work with for them are like large PDF documents right and these are not the types of data sources that have been valuable in the like you know structured data environments right but suddenly those organizations are like wow like actually the data that we've had like is suddenly really really valuable so a really good example is the IFRS um you guys I don't know how much of your audience will be familiar with the ifs they're the International Financial um regulatory standard Setter so basically you know in the US we have Gap accounting rules right um the IFRS set the sort of Gap equivalent for essentially the entirety of the rest of the world outside of the United States right and so they write all of these um you know accounting uh rules and regulations that pretty much every company every multinational company in the world has to comply with right and the data that they have is essentially the sort of financial reports of every company in the world out you know outside of the US but also frankly including us um that has you know ever been published which is a you vast amount of data but not the kind of data that you know the cloud native Tech Revolution was was really about like getting value out of organizations like that um also like you know biotech where you're reading a lot of research papers um you know all of a sudden there's a huge sve of data sources there that are extremely valuable um that historically have not actually necessarily been that valuable and the challenge is obviously those companies recognize the potential value but they're not as well positioned to adopt it today right and so there's much more of an education piece about like these are the sort of initial baby steps that you need to take in order to get it and you know we've been lucky to work with some like really Forward Thinking um okay so I was just saying uh you know these sorts of PDFs and like long text Data have historically not been that valuable now they're extremely valuable but um the big shift has been like Enterprises recognize that that data is extremely valuable now and we've been lucky enough to work with you know Enterprise it leaders and CI that um are really excited to adopt this technology because they recognize the value but also recognize that there's a lot of pre-work that they have to do before it's like actually unlocked right that's a really good point because in HR I mean we're all policies and we're like oh data oh you must mean the HR analytics team that has workday information we're like no actually all those policies is data like and the next question is data do what well you get my bot connected to it it can start answering questions and your customer service team is now relieved of having to answer all these garbage questions and then I start popping um so definitely great points raise so okay uh I want to uh this is how I came across you on LinkedIn you you had this let me see if I can get the LinkedIn get to work properly for once in my life I'm going to share my Fe feed here hopefully I don't dox myself nope not that one I want no not this one either give me a second here the glory of live uh production here no you had a uh you had a LinkedIn post that showed your growth trajectory which was incredible and I want to show everyone this real quick because um it just shows how kind of I mean it's so awesome you sounds like you found product Market fit here and so you said um actually why don't you explain this real quick you had thous 100,000% usage growth maybe you can go into that yeah so it's just a real number so we just hit like one year since launching the product in April it was in private beta but but the like we had that Enterprise customer the one I was talking about earlier um and it was like fully rolled out at the organization like as of April last year um so the end of April this year like we had our one year anniversary of of like you know having a publicly available company uh product that people could um you know could use and the experience of that growth has just been obviously like surreal I I couldn't conceive of being able to say that you know we have a thousand times more users and usage than um we did this time last year and I was already really happy honestly with where we were in April you know we had big customers that were household names that people had heard of using us um and I think one of the things that's just been really phenomenal is seeing how every time someone gets their hands on this technology they are like solving a problem but in solving the problem they're discovering 10 more problems that they could also solve right and so not only do we see this explosive usage growth from like new customers coming on but explosive usage grows from almost every single one of our users where they're like the more they use us the more like problems we unlock for them and the more um the more they can do with us and that I think that's like a really really powerful compounding effect where the the product just gets more and more valuable uh in people's hands so that's been really cool fantastic and um are the teams that are using this are they saying we're now saving X hours on approvals for stuff now or like who who been the biggest winners so far yeah so I mean obviously there's a ton of productivity savings that you get where you know it's folks like um you know the kyc team that I was talking earlier where they would do like roughly um a th000 on boardings like completed would would take them like roughly two months to do um and now a thousand on boardings can be done in like roughly a week um so yeah which is which is crazy and that's cuz like 92% of the time it's like a effectively instantaneous um decision right so that's like a really really really big deal um from a savings perspective but I think actually honestly as cool as those productivity gains are my favorite like stories are always the stories of things that were like literally not possible um previously that are now suddenly becoming possible because you know the data exists or um decisions getting made right so one of the things that I was speaking to like one of our we have a customer or end user really I should say he's a product manager of um at a big B2B SAS platform and uh she was saying like one of her challenges is that uh they deal with very sensitive data as well and so she basically uh the one of the big bottlenecks to their sale is how do we figure out like what security features we need for our customers to buy um and what's tricky is that um you know the conversations they have are mostly with the end users so they don't really get exposure to the buyer's security concerns and the buyer security pains and that was always coming secondhand through the SE the sellers who themselves don't necessarily understand the security considerations that well and so aren't like fantastic um you know they do a fantastic job of relaying the pain points so what this PM was able to do was take in all of the call transcripts that have been recorded across all of the sales conversations and suddenly start asking questions like which customers have complained about like the way that we store data in you know this subprocessor right and she can get the answer to that question in like one second in a way that before like she wouldn't even know who to ask right like who do I ask that question right um and so it's not as a result of that they were able to like re architect like the way that you know a particularly sensitive uh part of their platform worked and that actually unblocked this like massive deal where they were thinking about building it in house versus from buying from that customer and because of this like improved security posture they were like actually maybe we don't need to build an inhouse at all right and that just would never have happened without the ability to like figure out the information that you need at your fingertip tips without even knowing who to ask without knowing like where the information is just being able to get answers to very nebulous questions instantaneously um was like strictly impossible before that's huge um really um well said point one thing I struggled with at Google when we hit 187,000 employees you might have a question but you just don't know who to ask in your organization and you're just ping ponging from person to person to person I just give everyone PT TSD and also maybe ttsd flashbacks just you guys talking about it I can't tell you how many great great meetings ended everybody's so excited and then somebody says wait who has that data and then we all just realize like we're never doing this it can't be done now there's no answer to that question the moment you ask that question you know okay let's just go back to doing whatever we were doing before so true it's so true and um and then let's say you do have a question where we can get an answer you have to then go file it with maybe the data analytics team and they're going to get back to you in four to 6 weeks maybe or they're going to push back to you on this you're talking about Ravine is get your answer immediately no friction and that is really special so you were talking about the time savings for folks um let's go to really quick I like talking about pricing um oh I had a qu I had a question in the audience uh Daniel who is a contributor for our show Daniel thank you so much for contributing contributors um get access to our private Discord they get access to our live stream um past episodes and also AI research papers and people who people who are interview here such as Ravine not putting pressure on him but he gets free access to Discord Community maybe he can CH every now and then so um his question is could I connect your product to amplitude data yes absolutely yeah um so amplitude is uh like another really important source of information from a customer like I think one of the things that differentiates Credo is that we support both structured and unstructured data sources um so both unstructured data sources being things like images videos PDFs documents but structured data being things like you know your databases like your product analytics that kind of thing that that lives in amplitude so yeah a lot of our customers connect things like amplitude mix panel um you know segment is a is a big one as well like these sort of things are really really common for us awesome uh let's go to pricing it's always an art and art mixes with science isn't it so maybe just go into it love to hear your thoughts on how you came about with your pricing yeah if I'm honest this this pricing page kind of depresses me I I thanks Jordan the um the reality is like everything south of Enterprises like contributes like 2% of our Revenue right 3% of our Revenue so anything if I change the small companies teams plan to be $1 a month uh it I wouldn't really notice it on the um on the Enterprise on the pricing careful there's going to be a lot of smbs calling you up now I'll take I'll take that dollar sign me up for the dollar plan um so honestly for us like the pricing uh the value of the S&B Market is less actually on the revenue side but it is quite a valuable source of product signal um because especially snbs like they like to engage much more more deeply uh you know they'll answer your questions and slack they'll ask you questions and slack you get really high quality product feedback from them and obviously some of them grow into Enterprises over time and that that's great as well but for us like the pricing on the SNB side is very much just like how do we price in such a way that the customer like thinks that this is important right and cares about it enough to like do stuff to make it valuable for them um and you know the the revenue is obviously nice like I'm not going to say I don't want the revenue but it's like less critical for a business on the Enterprise side that's where essentially all of our actual money comes from and uh there we we have like a a much more sort of sophisticated Discovery process with the customer where we sit down and we really understand and learn and and figure out with them like what they actually need and I think like one of the things that people don't you know it's a notorious meme in like buyers of enterprise software that the sellers of enterprise software will just like never tell you what the price is right and it's really annoying and and I totally sympathize and empathized with it and when we started credle I was like I just want to have ridiculously transparent data like pricing and make it really obvious for everyone um and what we discovered pretty quickly is that the reality is that from the buyer perspective there's like especially at the Enterprise there's like eight different stakeholders that want to buy this thing and each person has a different perspective on you know what the the way the pricing should be structured and you know how to like tune the pricing such that the value that the company the customer gets is like commensurate with the price that they're paying and that's actually good for both customer so that they don't get ripped off but it's also good like for the company and the customer in in their partnership in the sense that if we're providing a product that's really really valuable for a you know particular subset of customers we can invest more in the product features and the delivery that they need versus the customers where okay like we can do a bit more of a light touch we provide a certain amount of value but it's not like you know we're not radically revolutionizing this company's business right and so actually the places that we're most valuable we can invest the most in like designing and reshaping the product for them and that's actually a win-win for both the customer and obviously for us as well so that's I think why we ended up at this like you know unfortunate like in like not super transparent Enterprise pricing but but I think that is actually the right thing not just for us but also for the customer as well because that's how they end up with the pricing that they actually want right um if we just had like a standard price for everyone then you know you'd never be able to keep all eight people happy um and actually at the end of the day like sometimes you just need to figure out a way for the pricing to to do that to get the deal done I mean Jonathan this is your area I would love to hear what your thoughts are on this sure as he was talking I was just I was again PTSD from my day selling enterprise software and uh the reason why enterprise software has to be priced differently for every customer is because they're Enterprises and they get to tell you whatever they want they get to say I want to buy per head or I want to buy it per department or I just want to pay a check at the end of the year or I want to pay per usage at the end of the year they get to tell you what they want through the 800lb gorilla and they expect you to accommodate them uh especially if there's a very large difference in size between you and them and what they're really used to is uh you telling them you're going to give everything for 50% off but then you're going to make up all the margin on Services because that's the way software and services used to be sold uh and they and the person who buys the software doesn't have to budget for the services that's somebody else's department and so they know you're just going to stick it to somebody else in the company so like this chicanery has just been going on since Oracle used to PL every one of their customers over and over again and it's just been a variation and and uh you know this is one of the reasons why you know smart people who started SAS companies started with SNB because they just can't and this is why you know even even in this day and age even when everything is like disintermediated companies who sell the Enterprise even like paler you know famously Peter teal made the made the case in zero to one you still need an army of salespeople you can call them a CEO you can call them whatever you want you can call them Consultants but somebody has to get in there and figure out what the business seeds and then figure out what you got and then adapt it right I think one of the things that so sha sankai is the CTO at paler um and I reported to him for a while while I was at uh at paler he told me something which I remember to this day like very acutely which was that uh this I think 2016 2017 he said I used to think this is him speaking right like I sham used to think that um Everybody hated Oracle and IBM because they were such bad businesses but now I realize that their customers hate Oracle and IBM because they're such good businesses right and um that was a very very interesting Insight now um I think you can take from that what you will but I think it came from a very deep Insight yeah I used to I mean I was in a business where I had to write checks to Oracle and we all swore whenever we wrote the check but we still wrote it and we could have written it somewhere else right but like you know yeah you got me in the memo you bastards well apparently we used to write like these comically gigantic checks at AWS right and you know you show up to the pricing negotiation with AWS and like they know that you're still going to pay them right and you know that you're still going to pay them and you go through this chicanery of like you ask for a discount and they know you're going to ask for a discount so they just inflate the price by 30% before so then they can be like oh no problem 30% discount because we're M no problems right but that the whole thing is like this elaborate dance to like keep all of the strange stakeholders happy but like the system does actually at the end of the day work because as much as pal was paying for AWS they were getting like you know that much value and more from it so I I'm sympathetic to the poor paler person every now and again they would try to get some leverage hey we're we're trying out Google Cloud this week the a person Just Smiles Winks let me know how that goes out for you sure it'll be great you'll be back you'll be back uh Robie first of all uh you didn't know me from Adam we didn't know each other and I just reached out you in LinkedIn I was like hey can you be on this podcast and you're like okay show'll me a doc about this okay I'll be on like thank you for the faith because this could have been an MLM or something thank you also because it's late right now you're in the UK correct so actually I'm in New York so okay a it's only four o' never mind let's do more let's do six more hours no uh um thank you so much for being on the show we'd actually like to have you back on again um actually want do one more question uh are you uh now that you you've found product Market fit and everything are you going through scaling where you're hiring your sales team out and everything are you fundraising um what's what's going on before you run I just lost I lost you for like five seconds you said now that oh yeah I was just I was just kind cussing basically the whole entire time no what I said is now that you found like your product Market fit are you now scaling up are you bringing in the sales team hiring are you fundraising what's next on deck for your company yeah we so we are scaling up we just hired um two folks to help out on the go to market motion which we're already excited about we're continuously hiring engineering um Engineers like um Gold Dust basically when you when you have a product that works uh just being able to scale it more and more effectively is like super super awesome um so yeah always always hiring on the engineering side fundraising right now to be honest with you like is not that big of a priority just because um it's harder to find fantastic talent that you're excited about working with than it is to like um like find customers that want to use you right like actually the bottleneck is like the bar of talent um that we're trying to maintain and so that is actually uh not honestly not the biggest priority we did a relatively chunky seed round like $4.8 Million last year and of that we still have like you know more than I think probably more than 90% of it still on the bank so yeah exactly just because it's easy it is to spend it when when the bar is right keep I mean well you know how to run your own business but like if you can continue just to like continue that that Financial discipline so that maybe this thing can eventually cash flow itself like that would be fantastic at the end of the day like you know if we raise money it will be because like we can invest in the company's growth like there's no point putting yourself in the position where you need to raise money to stay alive if you don't have to put yourself in that position right now obviously if the opportunity arises where we're like hey if we just massively increase our burn we can you know massively um speed the scale the scale of growth of the business then great but obviously you know that's one of those things where it's easy to persuade yourself that spending money will accelerate growth but it's actually much more difficult to genuinely translate like you know spending money into accelerated growth and so you know we are quite cautious and quite disciplined about do we really have conviction that like you know hiring another 25 people at like you know half a million dollars a year is actually going to generate the like commensurate like um increase in growth versus the inedible wor that we're already seeing you know exactly let's be real managing people sucks yeah people people suck maning people is so hard and it's sort of funny because like great people don't want to be managed right so you can't really manage them anyway um and then obviously if you hire like people that like you know aren't great then it really really sucks and so like but the the thing is that you know as a organization gets larger the complexity of the organization the communication barriers and all that stuff become become much harder it's so nice that like the whole team today sits in one room in our office right we can all just like look at each other and like ask any question you want um you know you have no there's like literally zero barrier to communication at all and like that's such an amazing feeling I really want to preserve that feeling for like as long as possible hell yes I remember um I don't know Mark like personally but he mentioned like a decade ago he's like I'm getting all this pressure that I need to expand to IND with the head uh headquarters New York and everything else like no I like keeping everyone in Meno Park and keeping it small and not managing all the time loans time zones of course he gave on that but if you can hold on to that now don't let anyone take it from you anyways Ravine we're gonna have you back on again this was great I'll invite you to Discord uh Jonathan you crushed it thank you all for watching uh if you like if you like this episode just please make sure of of course you like this episode hit the join button to keep us keep us going and uh I'll have a link to Ravine company so you all can check out out talk to you all later peace

Channel: SVIC Podcast

Views: 495

Keywords:

Id: aGTqhpW85po

Length: 63min 59sec (3839 seconds)

Published: Fri May 24 2024