E310 - Shubh Sinha, CEO, Integral Privacy Technologies
[00:00] Debbie Reynolds: The personal views expressed by our podcast guests are their own and are not legal advice or official statements by their organizations.
[00:12] Hello, my name is Debbie Reynolds. They call me the Data Diva. This is the Data Diva Talks Privacy podcast where we discuss data privacy issues with industry leaders around the world with information that businesses need to know.
[00:24] Now I have a very special guest on the show. I've been a fan, I've been fangirling you for a long time on LinkedIn.
[00:31] Shubh Sinha, he is the CEO of Integral. Welcome.
[00:36] Shubh Sinha: Thank you for having me. Excited to chat today.
[00:40] Debbie Reynolds: Well, I've had the pleasure of collaborating with you all on a couple of different occasions and I'm really happy and proud of all the work that you're doing and that you're getting so much traction in the marketplace with what you're doing.
[00:52] So you all focus on technologies, privacy technologies.
[00:58] And the way that you work is to me is very fascinating because I feel like, and I want your thoughts as we get going.
[01:05] I feel like people don't understand the privacy implications of some of the work that they do in industry, in different industries and how that plays out. But I want your thoughts.
[01:18] Please introduce yourself, tell us a little bit about Integral and yeah, what got you interested? What made you feel like this is a problem that you needed to attack?
[01:29] Shubh Sinha: Yeah, yeah, definitely. Looking forward to chatting today.
[01:32] My background is in building call regulated data analytics systems. And so what that means is before my co founder and I started Integral,
[01:42] my past life was building complex analytics environments to analyze very sensitive data.
[01:47] Consumer data, media data, healthcare data for many of the Fortune 500s. And I was the best way to think, kind of think about it is I was the consumer of deidentified data sets in order to produce insights for the Fortune 500.
[02:01] These insights could lead to better R and D initiatives, inventory rollout initiatives, or even ad tracking. And so kind of all across the board in terms of use cases. But the core theme was putting together very sensitive consumer data sets that describe people like you and me in a very in depth way in order to produce some sort of business outcome.
[02:21] When you're working with super sensitive and oftentimes regulated data, there's oftentimes various sets of regulations that apply. Healthcare has, hipaa, financial data has its own, there are state regulations such as CCPA as well.
[02:33] And so working in a regulated or sensitive data environment for my whole life, I've oftentimes worked with independent consultants or solutions out there that provide de identification services to make sure a radioactive data set can become compliant and ready to engage for business purposes in my past life, I found that the solutions today are extremely manual.
[02:57] They oftentimes lead to destruction in the data, meaning a quality degradation. Which as an insights provider was incredibly important to me because if it's trash in, it can be trash out.
[03:07] And then oftentimes what we found is that it was also not operating at the speed that enterprises needed and that people needed for better products or better services.
[03:16] And so that kind of firsthand exposure gave me the idea to figure out how do we automate a lot of this while keeping the robustness and the security that DE advocation is meant to provide.
[03:27] And so integral was born out of that insight where my co founder and I initially developed the platform to automate a lot of healthcare de identification. So linking various patient and consumer data sets together in order to gain deep insights into people like you and me, even in your own life,
[03:42] you can think of your non medical life and your medical life in combination influence the outcomes, the treatments, et cetera. And so we initially focused entirely on healthcare. And as AI happened, a lot of the models started exhausting the public data sets, what's publicly available in research papers and the open Internet.
[03:59] And we found that as more real world data came into the models, whether it be code bases, CRM, supply chain logs,
[04:07] name the private data set you want to, we noticed the same obligations based de identification was necessary for those data types as well. And so we have not moved away from healthcare, but we've expanded our ability to process proprietary data.
[04:20] And so what we call ourselves now is the independent privacy layer for AI, where our goal is to take private proprietary data,
[04:26] transform it to make it compliant within regulations or contractual restrictions or whatever obligations apply, while also retaining the signal that enterprises and AI companies need in order to produce a business outcome.
[04:39] And we do this continuously, efficiently, with a mix of software and services.
[04:44] Debbie Reynolds: Yeah,
[04:45] right. I always thought that because your origin in healthcare helped you quite a bit, because a lot of what we're seeing now is that so much more of data is regulated than was before.
[05:01] Right. And the regulations are coming fast and furious. So being able to understand that data is really important. But tell me a little bit about just the proactive nature of some of this.
[05:13] So I feel like some companies,
[05:15] and I want your thoughts, I feel like some companies don't,
[05:21] or maybe in terms of privacy, some companies think, well, we'll just build a product and we'll do whatever and then we'll think about privacy last.
[05:29] And so I don't think that's a good way to go about it. But what are your thoughts?
[05:34] Shubh Sinha: Definitely have seen a lot of that. I think we call it like privacy at the edge in a way, where you develop the product or the service and then you figure out how to solve for privacy in the last mile.
[05:42] I think our take, of course, being in the privacy space is that it's actually easier if you build along with privacy versus treating it at the end. And a lot of examples we see are these complex data purchases that are particularly expensive and very sensitive data sets.
[05:58] And people across business teams will be planning out what they want to do with this data set, what outcomes they want,
[06:03] all the things that should be planned out. And then they go through a compliance review and they find out, oh, I can only keep these fields in a data set.
[06:11] I can only have this particular part of it. And so that throws the entire planning sequence.
[06:16] It throws a wrench into it because you're not working with what you thought you might. And so it's a very classic case where if you're working without the privacy as like a parallel track in a way, it can affect the real business outcome.
[06:27] And then of course, it's a, it's a trust label as well. Some enterprises have never recovered from not working with data in the right way. And so the way we think about it is the proactive nature of privacy actually helps you plan better.
[06:39] Such that same example I just said, we work with them and they are able to understand what they'll have at the end of the process of a particular data purchase ahead of time.
[06:50] And so when they're planning,
[06:51] they're able to plan around what's going to be true 99.9% of the time.
[06:56] And that leads to more efficiency,
[06:58] better accuracy of the actual underlying outcome, and then also the robustness of being privacy first and compliant. Like, it's kind of a win, win, win on all, on all sides if you can treat it as a parallel path rather than this last mile loop.
[07:12] Debbie Reynolds: How do you talk to people who think of privacy,
[07:18] this work that you're doing, as more of a cost as opposed to something that's really going to enable them to accelerate their innovation?
[07:27] Shubh Sinha: Yeah, well, I like to tie it to the number one metric that any data team is really concerned about, which is data quality. Your cfo, your data analytics manager,
[07:38] the position,
[07:39] you name it. They're all concerned about what am I actually getting for this data purchase. And so privacy is often thought of as a cost to unlock that data purchase.
[07:47] But really going one step deeper, it's a scalpel to say what can you actually have in the data and how can you get it in the most compliant yet highly robust way?
[07:58] And so I try to transition the conversation very naturally to, hey, what comes out of Integral is, yes, it's compliant, it's robust, it has the security that you desire, but really, the teams that are going to be actioning upon it care about what variables they get.
[08:11] What's the bias in the data set? Like all these things that really affect whatever P and L or whatever product initiative that the data is being used for. And so I try to transition the conversation to privacy is a checkbox in many cases, for sure, but it has additional components,
[08:27] meaning it actually impacts what's in your data.
[08:29] And every data team, the number one question is, what's in my data? And so once I draw that link, I find that the conversation becomes quite collaborative in the sense of, hey, I need this outcome.
[08:40] How do we work backwards? And then that's where the conversation gets really fun as well.
[08:45] Debbie Reynolds: I want your thoughts about combining data sets. So I see this a lot. So whether companies acquire it, whether they have data from one department,
[08:58] another department,
[08:59] they want to combine together,
[09:01] sometimes that action happens without really assessing the fact that you may be creating more regulatory scrutiny or more risk in the way that you're handling data that maybe you didn't have before.
[09:15] And this is especially relevant in the AI age, where a lot of companies are doing things with data that they previously either could not do or would not have done in the past.
[09:26] Shubh Sinha: Yeah, yeah, definitely. I mean, my, my take is that a lot of the data sets going into AI and enterprise today,
[09:32] why is that happening? It's because folks want to understand how the real account of human reality exists in the world. And so, for example, in the healthcare side, it's helpful to know a point in time, like somebody saw this doctor and received this prescription.
[09:49] But you want to have the longitudinal journey. You want to understand why they went to that doctor. Is it because they were vacationing in that destination or is it because they live there?
[09:57] Does the income affect that particular doctor visit? Like all these additional variables, you can keep going kind of up the chain and you want the A to Z, but because the A to Z journey in that particular example can tell you why that patient is acting the way they are,
[10:12] or why that patient has the symptoms that they have or the prescriptions that they have. And ultimately companies want to know why with that data. That's why they're putting so many dollars and so many people on these data purchases and activating them and even in a non healthcare example,
[10:26] you know, something, something that we see in the company data world is you want to know how a particular engineer pushed code. And so you can look at the code base, but how did they think about that problem that typically happens in a Slack or Microsoft Teams when they're going back and forth with their team on,
[10:40] hey, I think about the problem this way is that right? And you have all these, call it additional behaviors that really give you the longitudinal journey. And so the point of a data purchase or the point of collecting data is to have as much longitudinality as you can.
[10:53] And that typically happens in combinations of data sets because rarely even in your own life is there one system that houses all your data. And so we see that as a reality of the world.
[11:04] You have to combine data and that means your risk will always grow as you're combining different systems data.
[11:10] And so how we think about it is that's table stakes, that's where it starts. And so we rarely even work with companies who have one data set. And I think it's primarily because there are very few companies who are just operating off of a single source of truth.
[11:23] And so I find that just coming in with that base mindset gives people comfort and confidence that we've been thinking about this space.
[11:29] And because that's a majority of the companies that we work with looking at this longitudinal journey, we typically come in with prepared scenarios around, hey, if you're looking at this, this is what it could result in.
[11:41] Or here are the risks of bringing in something like this. And so it's almost like the perk of the game footage in a way, having seen so much of it.
[11:49] But I definitely think, yes, the risk increases with each data set you bring on,
[11:52] but each dataset also just brings such a rich dimension, typically about a person that it's so incredibly valuable for the end outcome. That's a risk worth taking on. And as long as you do it, eyes wide open and then with a partner ideally that has seen it before, you're able to get the best of both worlds.
[12:09] Debbie Reynolds: Absolutely.
[12:11] Now, a lot of times when people think about data,
[12:15] companies think about data, a lot of times they're very focused on the provenance of data, like what they obtain, what they have, what they obtain, what they pull in.
[12:25] But they may not really have a good data story as it relates to the lineage. So sometimes data gets duplicated, it gets lost, broken up and put into different places.
[12:37] And the purpose for which the data was collected in the first place really needs to travel with the data. And I feel like that's a huge gap that most companies have where they're trying to do that maybe in a manual way and it's not very successful.
[12:52] But what are your thoughts there?
[12:54] Shubh Sinha: Yeah, yeah. I mean, data lineage, especially as data sets are procured faster and as there are more data providers, meaning more and more companies are monetizing the exhaust of their day to day business, or more and more folks are starting data companies,
[13:07] data lineage becomes quite complicated the more entrants there are into the space. And I firmly think it's a very positive thing that there's more and more entrants because that means more and more rich points on people like you and me that can be ingested and that is used for better accessibility,
[13:22] for products, better medications, all of that.
[13:24] The lineage problem, though, kind of expands or really compounds, I would say, as more and more data points appear. And so the way we see it is we almost see it as like a deployment style and a byproduct of Integral, where when we work with an enterprise customer, for example,
[13:39] when we receive their data from all their various suppliers, we oftentimes work with those suppliers as well for all their internal compliance, their internal data refinement.
[13:48] And so as Integral spreads to the ecosystem, it's almost like ports, if you think about the ocean, it's almost like ports that connect to each other. And so what that means though is that if the same software that is releasing the data is then, and that data is then being received by another instance of that same software,
[14:05] we've built in the capabilities to say, okay, went from this port to that port,
[14:09] and that is Lineage as a byproduct of adoption. And so our primary goal is to take the world's proprietary and private data and make it safe for enterprises and AI companies to be able to use.
[14:20] In doing that, we end up working with data or like the origin points of the data owners, the data sources, as well as data buyers. And so as we scale up our adoption, the lineage grows because the network grows.
[14:33] And so Lineage is definitely something worth solving for, especially as more entrants come in, which is why we've tried to be very configurable in how we deploy, where we can go on prem, we can accept data, but when we go on prem, we're able to establish those ports in a way that connect to each other.
[14:48] And so at scale, Integrals should be able to provide Lineage just as a byproduct of adoption. And so we have specific feature sets and like specific initiatives around it. But Largely we think about it as if there's one independent trusted third party,
[15:01] then that connection and that lineage happens as a byproduct of adoption.
[15:06] Debbie Reynolds: Very good, very good. What is the key differentiator for people who may not understand Integral and what you all do? What is your differentiator in the marketplace?
[15:18] Shubh Sinha: Yeah, maybe just to outline a typical de identification process or why dedication is even necessary. Right. There's so many data sets that are either bound by regulations at the federal or state level.
[15:28] So like HIPAA and CCPA are a couple examples of federal and state regulation.
[15:32] Or there's companies that are self regulating, meaning they're monetizing a particular data set that they have as an exhaust of running their business, such as their code base or their CRM.
[15:43] But it contains data on people,
[15:46] employees, and so you have to DE identify it to make sure you can safeguard privacy even if there's no regulation. These companies are typically saying, hey, I don't want to release my customers data.
[15:56] And so I want to self regulate in that way.
[15:58] And so that's why identification is necessary to make sure that privacy is safeguarded in accordance with either a self regulation or an actual regulation that is out in market.
[16:08] And so that's where the conversation starts. Typically then what happens is if you're not leveraging Integral and maybe you're leveraging like a human solutions provider, so human services and whatnot, they'll come in and take a look at your data.
[16:19] There's a lot of data, it's very unstructured data, typically images, messages,
[16:24] notes, audio, et cetera. It takes humans a long time to go through it just because it is very dense, it's very unstructured.
[16:29] And so you go weeks and weeks to calibrate on what is the risk in the data. And then it takes another set of time to how do I treat that risk.
[16:36] So the treatment perspective which is do we include audio, do we not do we mask the voice, do we truncate the zip code? Like those types of treatments that are aimed at reducing the privacy risk, but maximizing the value in the underlying data.
[16:48] And then it comes to actually executing that treatment,
[16:51] writing up any of the independent variable verifiable documentation that you would need.
[16:55] And so that's a typical process with a call it like a human services based software or sorry, human services based company versus a software where we still have humans in the loop for the robustness and the security that a human can provide, but we leverage software to do a lot of the things that a computer is much better.
[17:12] At doing than a human. And so that same process with us would be a number of days, if not even a day to calibrate on what the risk is in your data.
[17:19] Because our software can ingest the data set, it's mapped out risk in accordance with policies and all the work that we've done before. And it can assess risk very quickly and say, hey, here's what's almost like red, yellow, green.
[17:32] A human can have that conversation with another human to almost take a scalpel like approach to the treatment that takes a couple of days and then the execution of that treatment can take another couple of days.
[17:42] And so you end up working with a timeline of like a week or like 10 days as opposed to 20, 30 or 40 days, because you still keep the human in the loop for their expertise.
[17:52] We have the software doing all the crunching and the processing in between.
[17:56] And what the outcome is is a data set that is typically aligned with regulation or self regulation and the documentation which is a receipt of the activities conducted by the human and the software.
[18:07] And so the way we think about it is like it's an augmented process to ensure that the speed and the efficiency are there because it is table stakes to operate in this world very, very quickly, given how quickly technology is improving, but also keeping the robustness of folks who have contributed to privacy regulation,
[18:24] or folks who are very familiar with systems and can do that calibration on the ground.
[18:29] Debbie Reynolds: Very good.
[18:30] So as we're entering this AI age,
[18:33] we have not the AI is new, but I think what's new is the rapid pace of development in AI and also the rapid deployment of artificial intelligence in organizations. We've never seen organizations really capture this type of transformational technology within organizations.
[18:53] But so part of the message that you hear a lot is like, let's collect more data,
[18:58] more data is better, you know, big data, all those types of things. But we're talking about a lot is kind of data minimization and de identification. Why are those things important?
[19:10] Especially as we hear people probably say the opposite, where we need to collect so much more data. Why is this important in AI age?
[19:19] Shubh Sinha: I mean, I think about the generational technology shift that we're going through as really a shift in the accessibility of intelligence. And so these models, their primary value is the intelligence is very quick, the very high quality, or theoretically is increasing in quality with each release.
[19:35] And so what that means is that the processing, so the intelligence processing, it's at such a high clip compared to a bunch of humans in an analytics environment that was True, like, you know, five, 10 years ago, you now have so much intelligence accessible at your fingertips, what you feed into that is directly correlated with the enterprise value you can capture from that intelligence,
[19:56] meaning a good data set, or like a limited, call it like a bad data set. And like a limited data set, no matter how intelligent the model is, it can only execute that intelligence on the right amount of data.
[20:06] And so the whole kind of concept of bad data and bad results out a model can help you have better results from bad data. But ultimately, the best possible input into a model, the best possible input into intelligence, is to give it something very, very rich, very, very aligned data set that can then take advantage of that intelligence and get you a really good insight or a really good action out of it.
[20:30] And so the way I think about it is intelligence is just rapidly increasing in its ability to perform.
[20:36] So you have to set it up for success.
[20:38] And that's where more data, because a model can obviously take in much more data than a human can. So more data that's fit for purpose, aligned to your use case,
[20:46] can help you reap the maximum benefit of that intelligence. And so it's kind of like car and fuel in a way. Give it the best car in the world, but if you give it bad fuel, you're only going to go so far.
[20:55] And so that's how I kind of think about more data being collected, more unstructured data, more messy data. The intelligence is just amazing to be able to work with it and so feed it what it needs in order to get the outcome that you would desire.
[21:08] Debbie Reynolds: Yeah,
[21:08] very good.
[21:10] Well, I'm going to throw a scenario out at you. Let's say it's a big conglomerate multinational corporation.
[21:17] They have, let's say the company's been in business for 50 years. And so let's say they have 50, 50 years worth of customer data and stuff like that.
[21:24] For me, that frightens me. I'm like, well, are these people still your customers for 50 years and why you still have that information?
[21:31] But what types of things would you ask or you would think about for a company like that?
[21:39] Shubh Sinha: Definitely. I mean, I would say for a company like that, there's never been a better time to adopt the intelligence that is being innovated upon with a lot of the frontier companies.
[21:47] And so I'm assuming in this case that that corporation, you know, they're thinking about how do I leverage AI for the maximum benefit to my consumers and then my company.
[21:57] And so the way we think about it is you have this treasure trove of data.
[22:01] Some of it might be accurate, some of it might be inaccurate, but likely it captures a ton of behaviors that you want to analyze.
[22:07] And so Integral is a very critical step in the process to making sure that data can be deployed. We're not the only step. To be very clear, we always recommend like legal teams, compliance team, that you should be doing the work that happens outside of Integral.
[22:19] But let's just say that work is happening. What we go in and say is,
[22:23] hey, this is data on all your customers in this past 50 year history.
[22:28] Some of that you may have consent for. You should check, of course on your end. But from a de identification perspective, here are the regulations that you might be thinking about.
[22:36] Like if that corporation is based in California, is there a CCPA component to it? Depending on the data type it is. Is there a HIPAA component to it if it's a healthcare corporation?
[22:46] So the first step is really about calibration. Are you aware of what regulations or policies you might have to adhere to? Especially if you're like an international corporation? Regulations stack on top of each other.
[22:57] You can't just pick the best one or the favorite one. And so we kind of calibrate to here's what could apply. And then there's the actual data treatment itself. And so here's what applies.
[23:07] Our agents can understand policies and interpret them to say, here's how the data should be treated.
[23:12] And then it becomes a deployment into an AI model and whatever the use case is there. But in that particular situation,
[23:20] even if you technically quote unquote own all your data,
[23:23] it's not always all of your data. You need to go in and see what does it adhere to, whose data actually is it. And then those policies themselves have different treatments for how that data is collected.
[23:32] And so mapping out the calibration of how to treat the data and then going in and treating it, that first part is almost always the hardest part that are kind of like the privacy calibration of it all.
[23:43] Debbie Reynolds: Very good.
[23:44] Well, what about the end of life of data? So I feel like this is a huge problem that almost any company can have. And that's sort of you've reached the point where the data no longer has a high business value, but now it has a high risk.
[23:59] Right? Whether it be cyber or privacy,
[24:02] what types of things would you ask for? What do you think about for a company that maybe is hoarding information at the end of life, where maybe a separate or maybe they're taking on more risks maybe than they think they are by keeping like old legacy data that may cause an issue because of regulation or because of,
[24:26] you know, data breach and things like that.
[24:29] Shubh Sinha: Definitely. I mean, I think there's a couple ways to take that. I mean, one just something that we're seeing is that as the frontier models advance their capabilities, they're looking for more and more private data.
[24:38] And so if you're a corporation that's not working with your data actively, it's not that useful. It's, it's just call it like risk on your balance sheet as opposed to value.
[24:46] All these frontier labs will purchase that data in order to advance the next frontier of whatever it might be. So whether it's a medicine model, whether it's like a medicine treatment model, or whether it's a supply chain logistics model.
[24:57] And so if you're a legacy company in one of those spaces and you're not using your data, you might actually be sitting on revenue in many cases. And there's a very safe way to monetize that data so that you can maintain all the robustness of trust while also, you know,
[25:11] collecting top dollar for it and helping advance the capabilities.
[25:15] And some win, win because as those models get better, if you can use them in your business, you get to reap the benefits of that. And so there's that component of it.
[25:23] The other component is if you're sitting on a treasure trove of data that is not being actualized, then definitely looking into this is the best time ever to have this massive, massive kind of supply of data.
[25:35] And so it's a mindset shift. But it's, hey, this is very risky. But also typically the riskiest data sets have the most value in them. That's why they're regulated. And so what can you do to leverage what is this increasingly,
[25:46] very fast growing source of intelligence, AKA these models, and apply it to that data such that you don't necessarily decrease the risk, but you get the ROI from it, since you already have it.
[25:57] All of those options are better than just deleting your data is the best way to think about it. And so that's kind of how I pursue that scenario.
[26:05] Debbie Reynolds: So what's happening in the world right now in terms of data or privacy or anything in technology that's concerning you most?
[26:14] Shubh Sinha: I think what's interesting is that the most recent demand for data is definitely driven by the advancement of AI, so probably not a hot take there.
[26:24] And the demand for data has never been more global, I would say, like people just even thinking about you and me and you know, the people we know in our lives,
[26:32] people across continents, across cultures, they transact so differently in their day to day that the demand for data is so high. Because AI has the capability to understand humans across all those contexts, in every global context in the world and every company in the world.
[26:47] And so because there's that hunger, there's that capability, you have this massive amount of data excavation happening from all around the world. Some places it's regulated, some places it's not regulated.
[26:57] So you have this patchwork in a way of some data going in very robustly or like in a robust way, versus a lot of places don't necessarily like they're catching up with the regulation because technology is moving faster than the self governance layer.
[27:12] And so something I think that's interesting, that's developing, that is also like it's a concern and also an opportunity for privacy companies is the patchwork of what's happening.
[27:20] Because I think we're all aligned, that AI has these incredibly impactful positive qualities.
[27:26] And personally, the better AI gets, the more I get to use it to automate more of my life or introduce a higher quality of life. And I'm not unique in that it can help everybody.
[27:37] But what that patchwork and kind of, I guess the value chain to get there is like the patchwork of privacy in a way. How do you get that data in there as safely as possible?
[27:45] Because I think everybody wants to move fast in the right way. And so that's maybe the most interesting and concerning and maybe equally like an opportunity in that way.
[27:56] But the patchwork of privacy as it comes to data excavation is on a global level. You can imagine. I can't remember how many countries there are, but there's at least a large amount.
[28:05] Debbie Reynolds: What are your thoughts about the rapidly changing regulation, whether it be AI regulation in different countries like the European Union, and then also the patchwork of state regulations on AI and privacy.
[28:21] What would you say to a company that's just overwhelmed with all that and how you can help them?
[28:26] Shubh Sinha: I mean, I personally view regulation as it can be quite additive. Of course you want to make sure you don't overregulate as well. There's plenty of case studies where that's not been helpful.
[28:37] And so call it like the right standard of treatment for anything, especially data.
[28:42] It's helpful to have the right balance of trust and speed. AI moves very quickly and so you never want to hamper that speed. But also, can you move quickly the right.
[28:50] We believe you can. It's not an either or, it's a both.
[28:53] And so to companies navigating this landscape, I'd really, especially on a global level, I really look at where certain regulations look like others. I would say one advantage that we have coming into other spaces is that because we were so focused on HIPAA for a while,
[29:09] our technology was not a HIPAA technology. It was a technology that helped us align with hipaa. But we found that HIPAA set a really good precedent for many, many good things.
[29:18] Despite it maybe being an older regulation, it set the. Set a particular standard that so many other standards might have different language, might have different intentions,
[29:26] but they. There's like. It kind of rhymes in a way. And so something that I found is that people tend to come up with very similar protections all across the world because there is this idea of like, hey, this data is going this, this place.
[29:40] So in the case of data, it's like, data's going this place. Look for these things.
[29:43] It might look different in Europe than it does in America, and there might be more stringent regulations there versus here. But the idea is that it kind of rhymes in a way.
[29:51] And so look for the common patterns and see what you can almost like, standardize in between. Because not everybody's working in a silo. There is a ton of collaboration across the globe, and there's entire institutions, both for profit and nonprofit, set up for this.
[30:04] And so one thing I think people think is like, oh, there's so many regulations, there's so many different rules.
[30:10] While in quantity, they could be high.
[30:12] You can find a lot of commonalities and similarities in them. And that's how we've been able to approach and grow quickly as well, where we can take into account the uniqueness that is across regulations or across business preferences while also seeing the commonalities.
[30:25] Because there are a ton.
[30:27] Debbie Reynolds: I want your thoughts about silos.
[30:30] So I feel like some companies are like Santa's workshop,
[30:37] where everyone has their own little thing that they're doing, and they hope this magic toy comes out at the end.
[30:42] And so I feel like AI is definitely changing that. Where companies are enacting or deploying these types of technologies that have, like, a horizontal impact on organization. But how do you work with companies that are,
[31:01] like, very siloed? Because I feel like in the past, a lot of.
[31:04] Especially if they have regulated data,
[31:07] not just hipaa, but other regulations, where they felt like the best way for us to work with this data is to silo it and silo it from other people within the organization so that we don't, like, make a mistake or we have Internal data issue and things like that,
[31:25] where we're saying if we treat the data in a certain way and make sure that we're protecting the things that need extra protection, like sensitive data and things like that,
[31:36] then those insights can still flow to these other areas.
[31:41] Right. Without creating that risk. But what are your thoughts?
[31:44] Shubh Sinha: I mean, I think there's definitely call it good silos and bad silos. And so just kind of thinking about the historical reason for a silo or separation for environmental protections, those are really good intentioned efforts and good solutions with the technology at the time.
[31:58] I think what's happening right now is a lot of call it like a culture shift, company shift and tech shift is happening all at the same time. So something that we notice with teams is that there are fewer silos because they can rely on certain tooling to manage separations within even like a team of data scientists.
[32:16] And so you're having more and more collaboration in the real world, while also while maintaining the separation that you need at your computer or in a data lake, for example.
[32:25] So I think that one call it a cultural shift that's happening because of a generational technology being released. I think the other side of it is even if you do have a ton of silos, the rate of iteration within these different silos was historically a problem.
[32:39] Meaning one group could get it done very quickly, the other group took a long time to review,
[32:44] and then that delays the entire feedback loop. Let's just say you want to keep those silos exactly where they are now that there are compliant models or like privacy technologies that can help speed up the work you end up having, even in the case of silos, you end up having faster iterations within each silo.
[33:01] And assuming good faith efforts and collaboration, the entire feedback loop is getting a lot faster. And so the way I see it is there's a shift to fewer silos happening.
[33:09] And even in the places there are, if they're adopting the right technology and the right practices,
[33:14] the overall feedback loop relative to what it was in order to get like a product out there or in order to analyze some sort of result, it's happening a lot faster because the individual parts of a process are being sped up, which then streamlines the entire process itself.
[33:29] Debbie Reynolds: Very good. So tell me, why was it important for integral to lean into privacy?
[33:37] Shubh Sinha: Yeah, so I think the origination for the idea around working in the deification and privacy space was looking at.
[33:46] So when I worked at the previous company I did, I saw there's, we had regulated data divisions and we had non regulated data divisions. And I saw that every data scientist, every data analytics manager, every business use case, it demanded speed, efficiency, high quality data.
[34:01] The differentiation between a regularly vertical and a non regulated vertical, or like a sensitive one and a non sensitive one,
[34:07] was purely the privacy of it all, which does not discount its value. In fact, it really elevates its value. And so everybody wants speed of data and they want the data quality.
[34:18] And so the differentiation was that privacy. And so if you could make that process maintain its robustness but move faster, you were feeding into the behavior that is necessary in order to power better products or better services, or analyze particular results.
[34:33] And so that's why we kind of see it as it's thought of as the biggest blocker yet we see it as the biggest accelerant to doing it the right way.
[34:41] And so that's where we're seeking to enable data for use case sensitive data for use cases in a way that almost makes it seem like it's regular non regulated data such that flows quickly, but it has all the robustness with it.
[34:54] That's the reason for the focus on privacy. We found that if we could make a difference in that world, we could allow the data to flow.
[35:01] And that's what people want at the end of the day.
[35:04] Debbie Reynolds: I love that you said that people do think of privacy as a blocker, but I think you think of it as a blocker if you have to wait to the end and then try to solve for it.
[35:16] But if you're doing it at the beginning, then you're off to the races, you can go much faster. It does accelerate your ability to turn around products and do different things.
[35:26] Because you thought about that to begin with.
[35:28] Shubh Sinha: Yeah,
[35:29] yeah, yeah, definitely. It's truly an accelerant in the right ways because the more people trust you with their data, the more data that you can collect for the models, as we discussed earlier, the models will perform better.
[35:39] And so we see privacy and trust as a real win win.
[35:43] Debbie Reynolds: Let's talk a little bit about trust. So this comes up a lot. Now we see it even more and more. I'm seeing a lot more and more in the news about consumer trust.
[35:55] Consumers trusting different companies with their data, companies trusting one another. And I feel like, to me, I feel like trust is the new goal.
[36:03] Right. And part of that is being able to handle someone's data in a definitely compliant way. But also I think it's above that. Right. So I think part of it is part of trust is not table stakes in my view.
[36:17] So I think A person that gives their data or a company gives their data to another person, they expect that you're going to be following regulations. Right. But I think that trust layer is just a step above saying but I want your thoughts.
[36:32] Shubh Sinha: Definitely. I mean, I think about the trust layer as almost like a multi layer process. So it's thought of as this very important single layer, but really it's composed of so many different layers in between where how you collect the data, how you treat it,
[36:44] what you do with that data. Integral, I like to think plays an important part in many of those layers, but we definitely don't encapsulate all of it.
[36:51] If you're a company that for example, doesn't have a compliance team, that's a layer that needs to be filled. That's a trust layer that needs to be filled. So we see it as this multi step process and where we can do our part, we try and make that part as efficient and as expedient as possible.
[37:07] Because the speed of business is quite fast today.
[37:11] But I almost think about it as, are you doing the right thing to create the trust layer at every level such that you can say, hey, look at all this that we're doing?
[37:19] Yes, it's sensitive, yes, it's regulated, yes, there's a really big risk associated with it, but we've offset it in these ways.
[37:27] And then as a result of that we have this kind of massive outcome. And ultimately the whole point of taking on risk is to make something better or do something better for people like you and me.
[37:38] And so I do see it as a, like a, almost like a bidirectional win win where I'm happy to let my data be used by a corporation if they're following the right rules and whatnot.
[37:49] Because I know I will receive the direct benefits of a better medication or a better way to distribute something or a better product or whatever the, the outcome is. And so I think when you really look at it as a trust based system where people give up things or like people use things to get other things,
[38:05] so to speak, it's this nice kind of exchange of value in a way that as long as you're doing it the right way, everybody wins.
[38:12] Debbie Reynolds: I agree with that.
[38:14] So if it were the world according to you, and we did everything that you said, what would be your wish for privacy anywhere in the world?
[38:22] Whether that be human behavior,
[38:24] regulation or technology?
[38:27] Shubh Sinha: Yeah, it seems it sounds a bit maybe opposite of how important privacy is. But I think if we, and then companies in other buckets of the privacy economy succeed.
[38:38] You almost think about privacy as a given. Like you don't even think about it because it's so cemented as this layer that's built in. And so obviously it takes a long time to get there.
[38:47] Where today in a car, you don't think about your seatbelts being effective because there's been so many years of seatbelts working and that precedent has been set. And so I think about it where we have a lot of work to do.
[38:58] We as in the privacy economy in a way. But the end result of that work is that we as consumers, because I'm on the other side of it as well as my data is being used,
[39:08] we like the end outcome would be we don't think about it because we know it's there, we can verify it. So if we want to go research how the data is being used, where it's being used, what the outcome is, all that is independently there.
[39:19] But the almost the best outcome is that I don't think about it because it's so well ingrained and it's not this last mile thing. It's baked in as part of the entire process from all the trust layers.
[39:31] Debbie Reynolds: I can agree with that. So privacy baked in, not at the last mile.
[39:37] Shubh Sinha: Yeah. Like the reason people think about it is because it's a step in the process as opposed to what if it were the process in a way?
[39:43] Debbie Reynolds: Yeah, totally. As a foundation.
[39:46] Yeah, I agree, I agree. Well, thank you so much. It's been a pleasure having you on the show and I love collaborating with you all because I feel like you have your head in the right space and you're doing a really tremendous work, not just with your clients, but also educating people about the foundational nature of privacy and how I can really accelerate what companies want to do.
[40:07] Shubh Sinha: Thank you for having me. It was a great conversation.
[40:10] Debbie Reynolds: Now, how should people get in touch with Integral?
[40:13] Shubh Sinha: Integral, yeah. Yeah. So a couple of ways. One, on our website we do have a reach out form and then hellosaineuril.com there's also a great way to say hello.
[40:23] Debbie Reynolds: Excellent, excellent. Well, thank you so much again. And now we'll talk soon.
[40:28] Shubh Sinha: Talk soon.
[40:29] Debbie Reynolds: All right.
