Latest / Entrepreneur Odyssey / Protecting people behind the data with Ben Ramhofer

Transcript

Protecting people behind the data with Ben Ramhofer · Entrepreneur Odyssey

• Tejneet Selhi: Hello everybody, welcome back to the entrepreneur odyssey This is the podcast where we sit down with founders and senior leaders who are building something that actually matters I'm your host Tejneet Selly and today I'm joined by a co-founder who spotted a problem hiding in plain sight inside some of the world's biggest enterprise systems and that's Ben Ramhoffer He's the CTO of Maya data privacy limited a deep tech company that has built Europe's leading cross-system data anonymization platform. Yes, it's a mouthful, but it's a very, very needed platform, and I'm really excited to delve into this in bit more detail. So if you've ever wondered how enterprises can safely use sensitive customer or patient data to train AI models without breaching GDPR or EU AI data acts, may it is precisely the answer to that question. Ben brings 19 years of experience building data products within SAP Data for Life and he's now channeling all of that into his own company that is ahead of the curve. Ben, welcome to the Entrepreneur Odyssey. It's great to have you with us today. Ben Ramhofer: Hey everyone, it's a pleasure to be here, so thanks for the invitation. Tejneet Selhi: Absolutely. Well I'm going to dive right into it because we've got so much to cover today and obviously you've spent a lot of time at SAP and I believe you had the idea for Maya while you were still there and I was wondering what was the moment you realised that this was a big enough problem where you had to leave and go build your own company around this idea? Ben Ramhofer: Yeah. So yes, SAP, saw a lot as well as saw a lot of data and even before SAP I was in consulting, so did client implementation projects. But especially at SAP, I've been a product manager and I had a development team and we built all different kinds of products, financial products, products in the health space as well. And it always was a problem having realistic data. whenever We built the software. It was me basically typing in some data into Excel sheets that could be used for training or for testing by the teams. And even though SAP had plenty of databases with all the data that we would need to do realistic testing, we weren't able to use the data because of privacy concerns. And that always was a problem. And especially in the last years at SAP, I was working in the healthcare sector. And there I did co-innovation projects with different hospitals. And in one of those projects, we talked about their test environment, how they test data, and basically they had production data in their test system. And that means like this is health information of real people and on the test systems, more people have access to the data. It's even partners having access. And this was the moment when I thought, okay, even Before I was always of the mindset like, okay, data protection, yeah, we don't need it. It's just like it's a burden and it's blocking us from developing good stuff. But that was the moment where it clicked and it's all about data protection. The word is wrong. It says data protection. It's not protecting the data. It's protecting the people behind the data. And I think this is really Tejneet Selhi: Yes. Ben Ramhofer: where it clicked for me. And then like in the journey following on with Data for Life, there the idea was to bring data from different health data silos together and give it to research and as well there we're dealing with the most sensitive data. So it became Tejneet Selhi: you Ben Ramhofer: obvious that this is something that you need to go into. And that was a time where as well I went out as a consultant, did implementation projects and a previous colleague of mine from Dublin, Aloka, who's our CEO now. He came to me, said that he already worked on the topic beginning 2017 at SAP and that he moved out of SAP and founded the company and asked me if I want to join in. And that was the moment where I said, okay, this certainly is a topic I'm really passionate about. things were, yeah, you can make a change. We need to protect the people behind the data. Tejneet Selhi: I love that, it's protecting people. I think that that's, and I really appreciate you and your co-founder in going on this journey, because it's much needed. And so for anyone listening who's never thought about data anonymization before, how do you explain what Maya does in a single sentence? And why have most companies still not solved this? Ben Ramhofer: Yeah. So basically what we do is we make data accessible, usable for companies while protecting the people behind the data. So basically you can imagine that every company they're collecting huge amounts of data by providing their services. So if you're, if you're going to a supermarket whatsoever, they're collecting a lot of the data, like they have data. which is only about the products that they have. So that's not personalized data, but they have data about the vendors, about the suppliers, about all different kinds of about the employees. So this is very personalized data and that data has been collected so that they can scale their services and provide good services. And most of the companies start with very little IT systems, but as they grow, more and more IT systems will be added. this is very important that they can use the data for themselves. So they have a production landscape, but because they need to make sure that that production landscape won't go down, they normally as well have a testing landscape and a development landscape to develop new stuff, to test stuff before it goes to production to make sure that the production system always will be on. Tejneet Selhi: Interesting, very, very interesting. again, so how does Maya, I mean, what if a client wants to implement Maya? How do you recommend that they do it within systems that they've had for many years, for example? Ben Ramhofer: Yeah. So basically this is all about that you first get a picture of, what is the system landscape and what are the use cases that you want to use? Is it only about like testing, but maybe as well you want to use the data for machine learning or AI training. There are other use cases where you need to take the data and give it to external partners to work with the data. And for all of that, it is very important that this data that you give out or that you use for testing, you're not able to identify who's the person in that data. So for example, if you have me, if you have Ben Ramoff in the data with my date of birth, with my address, that the data will be changed in a way that it won't be Ben Ramoff anymore, but Max Mustermann or Frank whatsoever with a different date of birth and that you are not able to re-identify. the person behind that particular data. Normally Tejneet Selhi: Thank Ben Ramhofer: we do that on a use case base. When we're dealing with clients, they have a use case in mind that they want to start with. Then we look into, what are the systems that you have in place? The good thing is, as we moved out of SAP, before we were at SAP, you only were able to scramble SAP systems. But after moving out of SAP, we extended the solution so that as well, it can be all different kind of databases, if it's Oracle, Postgres, you call it. So, now we can consistently change the data in all the systems. So that the BAN will be in the CRM system, it will be frank in the ERP system, it will be frank so that you still can do the test cases across all the systems. Tejneet Selhi: Okay, so it's scrambling the data, but it's using that same data across all of the systems so that, yeah, I guess it's, you can still test the use cases regardless of which system you're in. Okay, makes sense. Ben Ramhofer: Exactly, because I think what is very important there is like we are a lot of people are talking now about synthetic data. When I mentioned that I was a product manager at SAP, I created synthetic data. I just came up with data that can be used for testing. But Tejneet Selhi: Yeah. Ben Ramhofer: from my perspective, synthetic data is kind of a perfect world where you think about, OK, that might be a problem. So you bring that into the data so that it can be tested. But the real problems are in the realistic in your production data. Tejneet Selhi: Yes. Ben Ramhofer: maybe you have done a migration and you have like 20 validations that in one particular field, there can only be a number and not a string. But because of migrations and other things, it happened that there is actually a text where there should be a number. And then basically when you do your testing, when you run your production system, this is where really the system breaks. And this is what we keep. So we keep the data as it is in the system, but changing the identifiable information so that you don't know who's the person behind that, but the pattern, everything will stay in the data so that it's realistic. Tejneet Selhi: So coming from, so I've been at Oracle for many, years and I've worked with hundreds of clients, thousands of clients probably, where... This issue that you are solving for is the exact problem. It's using quote unquote real data, which we really shouldn't use in sandboxes and testing beds. But I think with what you're building, we can use real data, but it be scrambled up. So it's not identifiable. think that's, it's great. It's great. Ben Ramhofer: Exactly. And maybe as you said that as well, came across that problem when you have been with Oracle. So when Aloka reached out to me and told me about what he has built, I said, I didn't use a solution like that at SAP. But this problem is so old, there must be hundreds of solutions out there. Tejneet Selhi: Yeah. Ben Ramhofer: But if you start looking into what kind of solutions are out there, it is basically there are solutions where you need to upload all your data into the cloud. And if you're working with restricted industries like healthcare, they can't just upload the data into the cloud. Then you have those big software or even the KPMGs of this world that are doing customized projects to scramble you the data, but then a system will be very expensive and it doesn't scale. And then you have like point solutions where it's all about one solution detects PI, another solution is doing that or is doing the scrambling. And you have those solutions like SAP had some solutions for their own systems and Oracle for their own systems, but it's not cross the systems. And this is like where we came in to try to, okay, this is a solution you can build, you can deploy on site because in healthcare you're not allowed to. to use it as a cloud solution. So it's a containerized solution. The client can install it on site and then go with it. again, that is coming back to where I started. I thought that such a solution must be out there, but it was not. And therefore we build it. Tejneet Selhi: Excellent. So tell us a little bit more about some of the features of the system because I believe you have AppSafe, FileSafe, AI Safe. What does each one do and what would you say is the most quite popular or most needed, let's say, for customers? Ben Ramhofer: Yeah. So maybe as well more like a history of the company. So we started off with AppSafe. So this is basically like you have like three different systems and you want to do integration tests between those systems. So what we've done is we've created the solution that let's say you create a production copy, like you have your test system and there you have the production data. And then we go read from the database. We scramble the data and write it back. So with that, we as well keep the integrity of the data. So it is basically, you can still run the system with that scrambled data. So for example, we have a hospital in Norway that is training their employees. So they have a lot of fluctuation in the hospital and they're training new employees on the systems with scrambled data. So that was the first start of that. So test environments. But then we figured out actually that's not enough because like if you have automation or test automation, very often you have some files additionally. Let's say you have an invoice and on the invoice as well there's BenRammhofer on it. So therefore you need to make sure that if you scramble the system and BenRammhofer will be changed to MaxMustermann as well that will be done to the PDF file so that he can still run it to see if everything is working. So this is how then after having AppSafe we as well build FileSafe and they're like a use case. We build it together with the company. They do payslip generation, but they do it in India. So they give the payslip information to India, but GDPR, like you're not allowed to give that personalized information out of the EU. So what they do is they basically, they scramble the data, but they keep an identifier of the person. Then. the pay slips will be created in India and when they get the data back, they do a very simple search and replace of the data. So with that, they're compliant. And then after building that file safe, now everyone is talking about AI. We figured out that, and actually this is how the entire thing started. So Aloka at SAP, he worked on machine learning for. clients. That was in 2017 where it was SAP Leonardo. But the problem always was that the clients weren't able to give production data because those machine learning models had to be fine tuned. So the clients weren't able to get the model fine tuned. So therefore in testing everything looked great. But when the production data came in, the accuracy went dramatically down. And this is where our local started. OK, looking. into the topic and doing research at the Aarhus University. How can you use privacy enhancing technologies to change the data in a way that the patterns will remain? And yeah, this is basically Tejneet Selhi: and. Ben Ramhofer: where we started with the machine learning and now as AI is picking up, this is where we see a lot of traction right now because there are so many AI startups out there and they can only build their products on real data. And as well, like when they want to deal with a company and say, listen, you need to give us the data. Those companies would say, no, first of all, no. And second of all, we can't because of GDPR reason, things like that. you know, every time, even if you make a, the contract into a place that you're allowed to give the data out, but every time you give data out, it's a new possibility that the data is getting leaked or someone is losing his password and. the reputation damage that is being done to the company as well. It's huge. So therefore companies are quite eager not to work with those startups and with our solution now, those startups could get into the data. So that is our last product that is getting a lot of traction right now. Tejneet Selhi: Okay, okay, and yeah and in an AI world that we're living I think that makes sense, that definitely makes sense. And then let's talk about regulations because obviously there's the EU AI Act, there's GDPR, there's NIST 2, there's a wave of regulations and I feel like there's sometimes more coming out, right? Is that a tailwind for mayor or does more regulations also mean frictions in your cell cycle. Ben Ramhofer: No, it is, I would say it's a tailwind. So I think with GDPR, yes, it's already in place for almost 10 years now. And that is as well, it's kind of a problem because those companies, haven't had solutions in the past. So they continued on with using production data in their test systems. So they learned how to live with a headache. So from a sales perspective, it's now hard to get into the companies and getting dedicated budget for that because now the budget needs to be put on AI. But in general, those new regulations help quite a lot. If it's the EU AI Data Act, the UI AI Act, if it's in other countries, so for example, like we're looking into India where you now have the DPA. into Canada, into other countries where you have strict regulations because those countries are really good for us to go into because they're aware that it's worth protecting the data. Tejneet Selhi: Yeah, absolutely. And I suppose to some extent it's kind of stress testing your platforms. Right? Ben Ramhofer: It is, it is. And basically, like very importantly, we are built algorithms how to change data. And this is like where we look into new regulations, what kind of algorithms you need. Nevertheless, it is still like with a client how he wants to use the solution. So he can as well, he can say for one particular test case, he only wants to change, for example, the names, but the date of birth things like that will happen. So with that's crumbling, it doesn't the data isn't, per definition, like GDPR, anonymized, but it can be used for those particular test cases. And I think this is very important that we have that flexibility of the solution, because if you talk to anyone in that space, the thing is, if you anonymize everything plain, as well, all the patterns are gone. So it is very important to keep the balance between anonymity or how you change the data and utility of the data. And if you know the use case, how the data should be used for, then you can balance that. And for sure, like if you're thinking about selling your data, that's a completely different topic because as well then you need to make sure that there is absolutely new way to trace back to the people. So it really then depends on the use case. to make or summarize the answer, yes, those new regulations help quite a lot as well to... Yeah, improve the product with new algorithms. Tejneet Selhi: Absolutely. And you talked about use cases. mean, do you have different use cases for different industries? Or is it generally, it kind of sounds like the use case would fit across all industries. Am I misunderstanding or? Ben Ramhofer: No, is absolutely right. And that is an advantage, but a disadvantage together. Because if you can go to all the industries, you're a very generic solution. So therefore, it is very tempting just like to go wherever there is a client. But what we learned over the past years is you need to go into a specific industry because you need to build up your reputation. One client is talking to another client, they're at the same conferences, things like that. So it is very important to more like focus on the industry and then move on from there. Like what we saw for us, it's certainly like healthcare, finance and insurance, that are the industries that are like our first industries to go through because first of all, those or people working in the industries, they're aware how sensitive the data is. And as well, like there is a big need for them to use AI because like I can only talk about the German health system, but like if you see doctors spending 60 % of their time doing documentation, this is insane. So now we're working like as well with partners. So we work with another company on Nexus. they are building basically transcription of doctor-patient conversation. Then they de-identify the data, use an LLM to structure the data, and when it comes back, we re-identify it again, and then it can be written into the patient health record, which is increasing the documentation time for the doctors by a big amount. Tejneet Selhi: Sorry, decreasing the... Ben Ramhofer: The decreasing, yes, sorry, it's decreasing the... Tejneet Selhi: Yes. Yeah, yeah. That's I thought you meant. That's what I thought you meant. I mean, that's incredible. Just there, what you've said. And you've mentioned that Maya is a real deep tech, not just a UI on top of ChatGPT, for example. There are, of course, other LLMs out there. What does it actually take to build a product that anonymizes data consistently across... thousands of tables, whether it be SAP, Oracle, whatever. Yeah, how do you even do that? Have you got a huge development team? Are you utilizing AI? Ben Ramhofer: So it's both, have a good development team, but as well like we're making use of AI. So you already mentioned it. Let's take SAP because that's my background and all those systems. Or Tejneet Selhi: Close. Yeah. Ben Ramhofer: take an S4HANA system. So here we are talking about thousands of tables and like 10,000s of different fields. And that is just like the standard tables and fields. So most clients have then. custom fields, custom namespaces, things like that, and very often that is not very well documented. So what we do basically is we have our own PI detection that is connecting to the databases, is going through all the tables, detecting where is personalized information, and giving that information to the client, and already suggesting how to change the data. Because it's very important, like let's say in one table you change Ben. to Frank and in another table, Ben to Dirk, then the entire system will break because the integrity of the data isn't consistent anymore. So with that, we use PI detection and that is basically AI or more precise, it's more machine learning where we go into the systems to detect what kind of information there is and how to change that because. That would be, it's always nice in demos if you show how you scramble one or two tables, but if you're talking about real systems, yeah, you wouldn't want to do that manually. Tejneet Selhi: Yeah, you talk about thousands, sometimes millions of records, so absolutely not. Yeah, yeah. Ben Ramhofer: Absolutely, yes. So just giving an example, so we were just like did another project with a client and there it was within two weeks and that was the duration until the first scrambling of the system was done. That was with setting up the system, installing it, creating the connections and then scrambling all the tables in that particular system. Tejneet Selhi: Mmm, too wheat, that's... Ben Ramhofer: possible with just like manual work. So there we need AI. Tejneet Selhi: I've got so many questions running through my head because data has been a big core of my background actually and I've worked with so many companies that have very bad data and I mean just... you know, just the thousands of names for the UK, thousands of names for London. How do you cater for quite bad data or bad quality data or, know, yeah, things like that? There needs to Ben Ramhofer: So. Tejneet Selhi: be standardisation, I guess, across data before it enters Maya. Ben Ramhofer: Yeah, so the thing is there, we don't do anything on the data quality. Let's put it that way on purpose, because Tejneet Selhi: Okay. Ben Ramhofer: what we are trying to do or what we're doing is basically we scramble the data. But the thing is, if you have like bad data in the system, that bad data still should be there after the scrambling, because you want to use it for testing things like that. There are other tools that are looking into how to improve the data. But what Tejneet Selhi: Yeah. Ben Ramhofer: what we really focus on is keeping the scrambled data as realistic as possible. So therefore, if you have some bad data in one field, yes, that bad data will be there after the scrambling as well. Tejneet Selhi: I suppose that's the integrity of your tool. It doesn't matter what the data is. Your purpose is to anonymize the data and scramble it. Yeah, that makes sense. Okay, let's talk about we've touched on AI, but your product is fundamentally AI powered, of course, with great developers behind it. You use ML to scan systems, as I understand, and detect where personal data is hiding. How do you think about building AI into the core of a privacy product, given that AI itself at the moment is raising so many privacy questions. Ben Ramhofer: I think there I need to be very precise because otherwise people would interpret it. So we are using an LLM, but that LLM is deployed, not an LLM, use machine learning and that machine learning is deployed on-prem at the client side. So it's not that any data is getting out and it's not generative AI. So we are not creating or working with text, creating new text, things like that. That is basically as well going into pattern recognition. going to the system. So you don't even need to have GPU machines to run that. You can run it on CPU, which is way cheaper as well then because if it's done in the client side. But I'm fully with you. we couldn't use any public LLM because then the data would go there. That would basically Tejneet Selhi: Exactly. Ben Ramhofer: undermine whatever we do. Tejneet Selhi: Yeah, yeah, no, absolutely. And I asked that because I'm sure that question be on the minds of listeners and also probably on the minds of prospects, I would imagine. So I think you've answered that. You've answered that really well. And then how are you using AI outside of your product for your day to day operational business needs? Ben Ramhofer: Yeah, so basically I do a lot myself, so of building my own agent. I, last year, just like that's more on a personal note, I did a course and I'm building my own agents that is supporting me on marketing, on sales, basically on everything that we do internally. So basically we look and we are still a small company. So we look into, what are the tasks that we can do? Where do we need to have workflows? Because often you don't even need to have agents. You can just like manage that with a workflow, building it with NNN or whatsoever. And sometimes you're building your own agent, but we are making a lot of use of that as well. And there we are using our own product as well with AI safe to make sure that no personalized data that from our clients or our prospects are going into those models. Tejneet Selhi: Nice, nice, it sounds like you're using it every which way you can, which is great. And I wouldn't expect anything less from a technology company really. So let's talk a little bit about the future. Where do you imagine Mea to be in say three years time? Because I feel like this tool is so important and should be used across huge large enterprises as well as everyday smaller businesses. So what's your trajectory, what's your plans for the next few years? Ben Ramhofer: Yes, so I think most importantly, build up a better brand or not better brand, but a well known brand. Because at the minute, like it's with a lot of startups you struggle with that people don't know you. So Tejneet Selhi: Mm-hmm. Ben Ramhofer: even like when we're getting into SAP deals, very often then a solution has been mentioned that for example, is only dealing with that SAP system. And that is solving a particular problem, but it's not solving the... the general problem for a client. But nevertheless, because that brand is well known, the client is going with it and say, okay, we're looking to the other stuff later on. That's the kind of no one ever got fired for buying IBM, that kind of statement. So it's the kind of where I think it's very important to build out more reference clients within the next time, and then going to as well to more conferences. going there with the clients, our clients promoting us. I think that is super important and that in our core markets, which would be like Germany, Switzerland, Austria, yeah, the UK and Ireland, plus now we're expanding to Canada and India. So I think this is very important to get more references there and then learn from the clients, basically what are their needs. What do we need to build for the product to forward? Tejneet Selhi: Great, great, I love that. Learning from clients is critical, I think, for success. And then, you know, I'd love to talk a little bit about your challenges that you've faced, and I'm sure you've faced many, but what would you say, with new founders in mind, I guess, what is one of the biggest operational challenges that you have faced, and how have you overcome that? Ben Ramhofer: So from an operational perspective, it's the kind of startup mindset. So everyone is on everything. So I think this is basically like whenever, like as well that is coming back to the problem I mentioned earlier, that we are not that well known. So we need to focus on quality. So whenever as well we are in the client conversation, it can happen that CTO, CEO and our chief growth officer are all together in that particular client meeting. which on the one hand, that time could be used as well to do other stuff, but we need to show as well the presence, how important that is to go into the clients meeting. So I think from an operational perspective, it's really about, and I think as well, that was a learning over the last two years that we have like dedicated tasks that are being done by dedicated people because otherwise everyone is on everything. And I think this is something that can't scale and yeah, this is. where we learned quite a lot now as well AI agents help or at least workflows where you don't need to do everything yourself, but still it's the human in the loop. You will check the results and still that is a lot of work. Tejneet Selhi: Yeah, absolutely. Because AI, you know, as I've said in this podcast many times and to clients many times, AI is a tool. It can't replace anyone as a human. It can replace tasks, which then means that us as individuals focus on other things. But we still have to check the results because it's not, yeah. It's not perfect right now anyway at least. Ben Ramhofer: Absolutely, as well you want to keep on top of things. So I think Tejneet Selhi: Yeah. Ben Ramhofer: it's not only like, I think there are studies like with LinkedIn where people are just like posting the stuff where they don't even remember what they've posted. So I think as well, it's the kind of where that is a completely different topic, but from a psychologically perspective, how you should use AI. And as well, there are studies that if you're only using AI to put in Yeah, your query and then take the result and give it to someone else. You're really getting your brain is, yeah, slowing down Tejneet Selhi: Yeah. Ben Ramhofer: on stuff. So I think it's about how you use AI in a good way that is Tejneet Selhi: Yeah. Ben Ramhofer: improving you basically on the things that you do, but not replacing you. Tejneet Selhi: Exactly, exactly. And I've heard this from a few different people. AI, you know, what you've got to do is learn to use AI effectively so that you're not replaced, right? Anyone who's kind of steering clear from AI, unfortunately, it's not going anywhere. Like any other technical revolution, it's not going anywhere. So we all have to jump on board to some extent. Ben Ramhofer: Absolutely and to be very honest, I forget about that way too often We are all and I would consider you in as well there. We are all in an AI bubble So Tejneet Selhi: Yeah. Ben Ramhofer: there are so many people they don't use AI if you use and look into the usage like most of the people that use AI Use the free version and there is such a big difference of using a free version forget about that the data will be used for training and things like that from a data privacy perspective. But on the other hand, like other people that are using it on a professional base. And I think like the scissors, like the difference between people using it and people not caring about it at all is getting so huge. And now you see those companies that are built with a few employees like take lovable and those like how incredible fast they scaled like I don't know 200 million ARR within the first year with I don't know 50-100 Tejneet Selhi: You know. Ben Ramhofer: employees something like that it's Tejneet Selhi: Crazy. Ben Ramhofer: it's that speed that is possible but you need to you need to make good use of it and not Tejneet Selhi: Yeah. Ben Ramhofer: just like what as well I saw with a lot of clients especially enterprise clients where The CEO said we need more AI and then they tried all the different things, but it's not clear what is the use case, what is the ROI? Does it really make sense to optimize Tejneet Selhi: No, no, no. Ben Ramhofer: that and bringing AI agents in? So I think this is where everyone was following the hype, but it was not too much like about business decision. Does it make sense? How much does it cost? What is the benefit? Sorry, I'm... Tejneet Selhi: No, this is fascinating because I agree with you. think it's, know, AI is a tool as I just mentioned, but you've got to understand what your problems are first. You've got to understand what your pain points are first so you can then build a strategy that incorporates AI. Right? cause yeah, mean, listen, we both work in the tech industry. I think we've both got many colleagues and friends that have been impacted by various decisions because just for the reason that you mentioned, because everyone's seeing it as a buzzword and say, we should use AI for this. We should use AI for this. We should, but you've got to take a step back and think about, what is it? So anyways, I think we could talk about this for hours. But we're almost out of time, unfortunately. I'd really love you to tell our listeners how do they get in contact with you if they want to start working with Meir. Ben Ramhofer: Yeah, absolutely. first of all, www.mayadataprivacy.com. So please visit us there. But as well, like you see my name, you find me on LinkedIn. I'm very happy to link up and have conversations and help you to make your data actionable for the company. So I think we gathered so much data over the past decades. And I think that data should be used in a good way. and we should make sure that it's not exposing anyone, not us, not our children, not anyone. So we should protect it. So please reach out directly to me and I'm happy to go into conversations and check how we can make your data actionable. Tejneet Selhi: Excellent. Well, look, Ben, this has been an absolute pleasure and a really fascinating conversation. I think founders who are just starting out could learn a huge amount from what you've built. And I do encourage everybody to get in contact with Ben Ramhoff, especially if you're building out technology, right? Or if you're even using multiple systems. think, you know, privacy is where you want to go to help you. you with scrambling data or anonymizing. So thank you again, Ben. It's been a real pleasure. And to everyone listening, if this episode gave you something useful, please share it with a founder who needs to hear it. And we'll see you on the next episode of Entrepreneur Odyssey. And be sure to please give us a follow. Thank you, everyone. Until next time.