SRUJAN AKULA - THE BASE LAYER OF AI: IT'S THE DATA

Summary

Highlights

• Data quality underpins effective AI – bad data leads to bad outcomes 🤔.
• Enterprises can cut data prep from a year to minutes with the right tools ⏱️.
• The rise of AI tools like chatGPT emphasizes the need for clean data 🧼.
• Modern Data offers a platform to make enterprise data user-friendly and efficient 💡.
• Real business changes occur when data is treated as a trusted product 📊.
• Simplicity in data access can unleash potential in sales and marketing teams 🛒.

Key Takeaways

• Data is the base layer of AI magic—it has to be high quality and abundant 🧙‍♂️.
• Quality data management can transform enterprises into data-driven powerhouses 🚀.
• Businesses need simplified tools to use data without tech hassles 🛠️.
• Treating data as a product streamlines AI implementations and boosts decision making 🎯.
• Modern Data Company aims to make data easily accessible and trusted for business users 🤝.
• AI's future lies in integrating quality data and operational simplicity 🧠.

Overview

SRUJAN AKULA - THE BASE LAYER OF AI: IT'S THE DATA Transcription

• 00:00 - 00:30 [Music] hi welcome to AI impact and strategy in our previous video we interviewed Devan ashar and she showed us how she uses mid Journey AI to create amazing graphics and to do it really really quickly in today's video we're going to go to the other end of the spectrum and look at the data which lies at the base layer of all AI efforts of analytics and why is
• 00:30 - 01:00 this important well a recent report from MIT technology review makes it pretty clear the report is about data and how organizations can become data driven using Ai and the very first paragraph of the report reads like this it says an organization's ability to generate actionable insights from data often in real time is of the highest strategic
• 01:00 - 01:30 importance to help them become datadriven companies are deploying increasingly Advanced cloud-based Technologies including analytics tools with ML capabilities and then the punchline here what these tools deliver however will be of limited value without abundant high quality and easily accessible data so the obvious key question here is how do organizations assure that their data is above abundant high quality and
• 01:30 - 02:00 trustable and easily accessible well to answer these questions today we're going to interview the co-founder and CEO of the modern data company modern is a Silicon Valley venture-backed startup company that is attracting a lot of attention and now a lot of customers precisely because of its unique approach to data data management and AI obviously modern has a lot to say about Ai and
• 02:00 - 02:30 data and uh so tell us the story please awesome so so Lang the mission with which we started the modern data company is to truly transform how Enterprise business users use data as a competitive Advantage when you see the market today every approach to data hinges on having serious technical expertise to be able to prepare the data and then think about what sort of AI modeling do I need to apply what sort of sites can I generate
• 02:30 - 03:00 to drive what sort of business decisions and that is not working out when you see the history of Technology the sooner we can abstract out all of this complexity the more mainstream that technology becomes and I feel data is at that Tipping Point where organizations are starting to see data as a true business asset that should be in the hands of the business users and give them simplified tools to drive that competitive advantage that's not happening today and
• 03:00 - 03:30 our mission is to make Enterprise data as easy as a business user finding it trusting what they find and being able to use it themselves without needing technology so that's broadly what we're trying to do at modern and so what's the product or service offering that delivers that on that promise yeah so we have a a platform that we call a data operating system the name of the platform is data OS and the value proposition is that we can deploy this platform on top of a
• 03:30 - 04:00 customer existing data infrastructure one simple integration into their storage layer converts all of that technical you know jargon around data into a very simplified business asset where security trust where did the data come from is this compliant with gdpr CCPA you know am I governing this right am I protecting my data all happens from this business asset which we call a data product we deploy within the customers
• 04:00 - 04:30 uh data environment data infrastructure so we are a very open platform everything that the customer does they have control over and it coexists and interoperates with any technology that the customer uses today to run the data inent so now how does all that relate to AI yeah great question and you're seeing a lot of probably um Investments being made in data especially with the you know the rise of chat GPT and the value that organizations have suddenly seen
• 04:30 - 05:00 with this kind of a technology in a broader internet knowledge carpus when it comes to applying that same technology to your Enterprise data there's a lot of challenges and most of these challenges lay at the data infrastructure level am I preparing the data in the right way so that you can run effective AI unless the quality of the data is good what you do with AI is going to reflect that underlying quality sometimes it could be dangerous if you don't fix and understand what is the
• 05:00 - 05:30 quality of your data are you protecting the data well are you introducing biases which might lead you to the wrong business decisions all of that needs to be taken care of with the infrastructure layer and that is where you know a platform like this comes into picture where we are inherently ensuring that the trust the quality the compliance the governance aspects of data are taken care of so that the business user can start thinking about what sort of models
• 05:30 - 06:00 do I want to bring in to streamline uh you know what sort of business operations and how quickly can I react to my business needs with predictive capabilities all of that hinges on high quality data which is trusted compliant and the business has the peace of mind that I can start building my models on top of this today if there are $100 being invested in AI I would assume $70 to 80 are going towards fixing the data versus investing in Improv improving the model improving
• 06:00 - 06:30 your predictions and that is the dichotomy I see where you know the most of the value is on the consumption side most of the Investments are happening on the management side and our mission is to flip that Paradigm okay so can you describe um some of the PE some of the organizations that are using um modern and how they're getting benefit from it in conjunction with AI so our laser focus has been on global 2,000 companies you know this large ENT price clients
• 06:30 - 07:00 that will see a massive impact if they get their data management right okay and kind of platform we provide especially allows large Enterprises to manage their data in a significantly simpler manner than what they're able to do but more importantly create these business asset data cataloges we call them data product cataloges that completely transform how Enterprises work with data just to give you some examples before technology like
• 07:00 - 07:30 data o from the modern data company our customers used to take maybe a year to build some new applications that will allow their sales teams to understand what is the next best action they should be taking when they visit a customer site as an example right so you said a year a year okay 12 months 365 days and yes it might sound crazy but it's multiple things it's just not technology it's how data teams are organized today how much dependence you have on your
• 07:30 - 08:00 technical tooling what sort of maturities is on the data side on the business side when it comes to data all of these add up to that yearh with our capability the same business teams without thinking about technology teams supporting them are able to execute those use cases build applications in streamlit react Etc to hand it to their sales associates all of this is happening within minutes versus waiting a year that is the transformative power of treating data as a business asset
• 08:00 - 08:30 okay well that's pretty interesting I've a couple of the other interviews that I've done on this channel uh people have talked about a 20x productivity Improvement but if we go we're talking about minutes to years we're talking about hundreds of X here I I believe so you know almost think of it like you know you're creating an app store for your Enterprise data assets which are findable inherently trustable right all to do is start using them with the Integrations that are already rebuilt
• 08:30 - 09:00 that is where the the one year becomes minutes and self serve more importantly and what we've also seen is given that this level of Simplicity to access complex data from a business side exists we are seeing like for example marketing managers be a lot more value driving functions than just executing campaigns and Reporting on the results they're able to get more deeper into the data to understand where is this resonating with what segment of customers how should I I change the messaging while the campaign
• 09:00 - 09:30 is running so that I can improve my Roi so functions that used to be more support functions are now becoming more value driving functions just because we have unlocked access to data in a much more simpler Manner and do the people who are doing that who are creating that value are they specifically using AI or they're just using analytical capabilities based on the data products that you're providing it's it's an all the above you know when you think about you know how do I streamline my sales
• 09:30 - 10:00 operations or marketing operations or my finance operations it is a combination of multiple things there's still a need for very detailed dashboards where you can drive these insights and understand what is happening you definitely need AI ml to do add more predictive capabilities and now with Gen becoming a major part of the organization's data strategy that also is a big uh value at how do I provide more simple conversational interfaces for my trins
• 10:00 - 10:30 to be able to understand what is happening and more importantly react very very quickly at the speed of your business so how many years does it take to install data OS how many years oh so we we can get a customer up and running in a few hours you know we typically deploy within the customers's cloud environment so all we need is the infrastructure provisioning to be done by the customer we have an automated way of deploying the entire capability within their azure WS or Google
• 10:30 - 11:00 environments and there is a fair amount of machine learning and automation which helps us understand the state of your data the relationships of your data how good is your data which source systems need Improvement what data is sensitive but not protected these sorts of things are fairly automated and once you have the automation then you go understand what is the business intent what is it that the business is trying to do how performing does this need to be how one does this need to be with respect to
• 11:00 - 11:30 data privacy and using that business intent we look at all of this understanding that we have and quickly map your current state into what you need that's sort of the magic that we provide as our platform so that's that means that you're actually applying AI techniques and Technologies to their data to enable them to then have better data so they can apply AI to problem solving is that correct 100% you you got it right I think there's a a big opportunity to use AI to make this
• 11:30 - 12:00 access simple so can you describe how you're using AI to help you manage their data so multiple things you know we are working with a large uh uh device manufacturing company which is using AI on their applying AI to their device Telemetry data to be better prepared for warranty management to do much better predictive maintenance of different parts that might you know go wrong within that device just one example a
• 12:00 - 12:30 simple example of how you know uh quickly converting that Telemetry data coming from multiple iot devices into Data products allowed the customer to now apply all of these AI techniques to to do better warranty management and better predictive M that's one area on the business side as an example I can so hang on so they have they're sending iot data to their cloud data infrastructure yes how are how are you using AI to help
• 12:30 - 13:00 them manage the data that's coming in that's my question when you getting data from all these iot sensors right sometimes data might come in there is a rate at which the events come in a lot of times if you know you you see problems with your sensors itself the rate at which the data comes in could be wrong sometimes the data coming out of these things could be corrupted so the first step is to understand the data that I'm getting from all of these sensors is this good quality data does this have enough coverage for me to do
• 13:00 - 13:30 maintenance algorithm on top of the data okay and that aspect of what should be the quality of the data what level of errors are okay for me to predict what I need to predict with the level of confidence is what a technology like what we built and a construct of data products provides you the fundamental essence of treating data as a product is you clearly state I'm okay with this quality to execute this use case I need
• 13:30 - 14:00 this data to be in this format so I can use this kind of ml model ml platform to run my you know models for example all of that is captured ahead of time so when you start using now a data product that captured all of this iot data the problems I told you just now about the device data quality and address out of the box how the data is protected is out of the box how it should be used and delivered is also out of the box that is where the uh applying these kind of data
• 14:00 - 14:30 as product techniques significantly streamlines your AI and shifts your focus to Value delivery so that's that's very helpful and clear um let's shift a little bit and talk about modern as a company just so that people have some context around that so modern is a Silicon Valley startup Venture back and you're going to completely disrupt the industry and I I truly believe data is at that c of becoming truly mainstream what I mean
• 14:30 - 15:00 by that is you know like think about as an analogy uh the mobile app store conert smartphone industry right before the IOS app store and the Android App Store it was a nightmare for a developer to build a mobile application test it against multiple screens multiple you know uh uh regions Etc how do I distribute that how does a user find my application and deploy it on their device was all extremely complicated iOS
• 15:00 - 15:30 came in and said why do that I'll standardize with an operating system I'll give you all of these apis and interfaces you just come in you have your intent in mind with that intent use all of my interfaces build deploy start monetizing instantly that is the same thing data is ready for that kind of A disruption where a business comes into these data product cataloges data product hubs finds what they want implicitly trust and start using that is the
• 15:30 - 16:00 transformative nature of what we are trying to do and if we are successful that is how I think data truly becomes a new oil where you know everybody from your marketing manager to your Finance consultant to your sales exec can apply AI gen in their business functions and that's the vision toward which we are working okay so how long is it going to take for that Vision to be really really widespread in mainstream
• 16:00 - 16:30 so we've recently announced our major product upgrade of you know a industry first business data ready data product catalog where you're not thinking pipelines you're not thinking data sets joints and all this and truly treat data like a product pick it up and start using that is already transforming like I said you know organizations that were thinking about a year to execute a use case are now able to do it self serve you know you know imagine a large
• 16:30 - 17:00 organization that works with thousands of suppliers that sends data to them through Excel spreadsheets today overnight to api5 that entire data sharing with third parties using trusted API layers with full security governance built which is allowing them to think about the monetization very differently than just sharing Excel files over Excel files with the customer so and these Transformations are happening within months you know versus multiple years
• 17:00 - 17:30 that it used to take so we're talking about a major disruption yes and we are already ready for that the platform is ready we are seeing a big uh momentum from large Enterprise clients that see this as a missing piece to close that last mile problem that they have in data how long have you been developing this the company has been around for about 5 and a half years now you know we've been probably heads down building this tech for a good majority of time and over the
• 17:30 - 18:00 last few quarters we've been really actively you know um scaling our sales efforts and seeing a ton of momentum said the next big innovation in data is going to come from an organization that can streamline the consumption of data make it simple for business kind of like what we talking about not from storing your data more efficiently processing the data more efficiently because that over the last 10 years has measured enough where it can it is starting to get get commoditized
• 18:00 - 18:30 okay so the focus of the work that we're doing on this channel is on AI so I'm interested in your thoughts about where you think it's headed in I don't know say two to three years both in terms of the data and in terms of of AI and and the evolution and Adoption of both the the approach that you're talking about prioritizing data and um the expansion of AI and it's sort of taking over the economy yeah so few things langon one we
• 18:30 - 19:00 absolutely have to solve the data management and making data ready for AI problem MH and when you treat data as a product forget about which te underlying platform you're using to make that happen it elegantly solves both sides it streamlines your data management and making the data ready at that base foundational layer and it also gives you simplified ways of deploying that data towards your decisions because all of the Integrations everything that you
• 19:00 - 19:30 need are out of the box available so I personally truly believe that treating data as a product will become essential to scale AI the way organizations are envisioning and when think about gen and what I tell our customers is there is I feel this I fear these inflated expectations where you know I'm just going to bring and gen model and magically you know everything is going to get sorted I think that is going to happen we are probably a year year and a half away from an organization truly
• 19:30 - 20:00 leveraging all of these outputs out of the Gen and AI models and automate your business operations or business decision what we are seeing right now is a lot of these techniques being applied to streamline your operational aspects of your business how do I streamline my sales operations better you know how do I streamline my cloud Financial operations how do I streamline marketing Ops that is where you know we are encouraging our customers to start playing with some of these techn Oles because there's an immediate win in
• 20:00 - 20:30 terms of the efficiencies you can bring from an operational side to these Enterprises and as these models mature you will see a lot of Automated Business decisioning happening but that'll only happen when you can inherently and implicitly trust the underlying data which is the biggest challenge that everybody is facing today we're getting revolutions now compounding on other revolutions which is driving this process of exponential change it's very interesting to think about again the base layer like you know
• 20:30 - 21:00 earlier you talked about you know if your dat if your data doesn't have good quality then you can't trust AI to run on your data because you don't know what you're going to get which is sort of a restatement a modern statement of the old saying garbage in garbage out right yes that's yeah that's kind of fundamentally you put it sly yeah that is literally what we are solving here right and it's exciting you know what we are trying to scratch the surface with what's happening with AI and gen today
• 21:00 - 21:30 and what I see happening in the future imagine specific agents that are looking at your proposals that are coming in to see what is the effectiveness of this how am I you know uh keeping an eye on my AA on my margins you know how am I uh serving my customers better with uh you know much more personalized experience I see these sort of Agents being built specifically to power and streamline each of these business functions and
• 21:30 - 22:00 eventually all these agents talking to each other to tr automate your business decisioning sales starting to talk a little bit about it with their agent force uh architecture where they want all these agents that are solving specific business problems MH I think that's a right approach but it again like we said it hinges on The Good the right data prepaid against those business intents MH as these data products which can power that agentic flow you know very very effectively this is not a generic problem like the
• 22:00 - 22:30 internet knowledge Corpus that I'm trying to learn on every business is so unique how their data is prepared what it means to their business is so unique it requires a lot more of you know what I just said one siiz fits-all strategy our agents now talking with other agents and we have a chain of agents that are either making decisions or making recommendations or you know providing input or is it one agent that interacts
• 22:30 - 23:00 with a human would then leads to another agent so in other words if there's a chain of agents that are working autonomously then you can Envision that entire business processes that are quite complex and significant financial cost implication and value um can be entirely autonomous and I'm just wondering how close we are to to that sort of a situation I believe that that is the right vision I think we are probably a couple of years away from having all of
• 23:00 - 23:30 these agents talk to each other because said organizations are still spending most of their money fixing their underlying infrastructure right and that stops and you start focusing on how this agents play well with each other how do I create these agentic flows that is where the conversation is to shift yeah when I see the market we are starting to apply these agents on specific domains the the way you have described all these agents being able to talk to each other towards a business automation I think we
• 23:30 - 24:00 are one and a half to two years away from that okay but it's coming but it is coming I mean I think it'll be the future I you've already seen it with you know process Automation and you know other areas which got significant uh what do you say uh value uh you know um realization to happen quickly because of the kind of automation we've done so what we're talking about is not like a new paradigm for SE that has already worked in the overall technology space
• 24:00 - 24:30 this kind of Automation and I data is ready for that I think two years it would be my guess well it will be a paradigm shift as the the stakes get higher like if agents are not right now working at a relatively low level logistically in you know organizing to solve you know FedEx or UPS delivery problems or you know things of that nature that's one thing but if agents are creating and feeding decisions to each other that are leading to business consequences then that's a that's a
• 24:30 - 25:00 whole different world the the general evolution of Technology implies that that's where we're headed which becomes a very very interesting sort of futurist perspective and the question is how soon that future is arriving and you know if you're right that it's a couple years then that's that's really pretty soon I think so I mean I yeah I I do believe this starting to catch fire because other thing we didn't talk about besides the quality uh issues is the data silos okay all this
• 25:00 - 25:30 data imagine if all of your e-commerce data sits in a silo and the data around how you're managing your distribution centers sits in a different uh Silo how do you now identify through an agent that this is the increase in demand I've seen you know in this store so a distribution center proactively ensure that you know the demand is fulfilled MH there is a little bit of that which needs to be fixed as well you know where the data talks to you know is not siloed
• 25:30 - 26:00 and it's more a unified view across Enterprise yep so that's that's available now or that's coming it is happening it's there is enough technology but I feel the customers that I work with they still struggle with some of those data silos where you know they haven't completely solved it on their end yet so in the in the data Lake model which is data Os or modern have a role to play in in that infrastructure as well uh that is exactly where we play is we'll sit on top of your lake or we'll sit on top of your Cloud warehouse
• 26:00 - 26:30 and make all of that Data Business ready from a consumption perspective any other thoughts or comments if you try to fix data using a fiveyear 10 year old approach you're are going to struggle if you are living in the era of AI geni there are much simpler much better ways of H managing all of your data needs so that that trusted base layer happens much faster than what you think it could
• 26:30 - 27:00 and if you look at what Gartner is talking about with data as a product why that category suddenly has become such a hot category in data is because of the Simplicity that it brings in its approach to getting data to be set up well from a foundational perspective how do you describe the end state in like two sentences I'll give you an example you know you have a a category manager in a marketing Department that is trying to understand
• 27:00 - 27:30 the propensity of your customer to specific categories and who should I push what categories to this is a fairly common problem that exists in a lot of Enterprises today okay this new world that I'm talking about you could literally bring a few of the Shelf models slightly tune them and as a business team try this out yourself and then then start applying it into your uh business decisioning a think that is exciting for us is I'm starting to see that way of just finding what you want
• 27:30 - 28:00 trusting what you want and using it instantly and not just doing it one time but being able to repeat it for every data access that you have to execute every use case and as a result of all of this data being available we have seen significant improvements in uh your campaign uh goals that objectives that you have set up we are able to see some of our distribution companies being able to uh fulfill campaign objectives you know for their end customers in a lot L better more efficient manner we have
• 28:00 - 28:30 seen significant uh Improvement in driving further sales for some of our customers we've helped improve marketing efficiency significantly we've helped e-commerce uh click through rates on purchases improving as a result of this underlying platform so in one of our customers at in the first 3 months of deploying data overs 40% increase in uh sales for their higher selling sales category because of the intelligence we are able to provide in being able to
• 28:30 - 29:00 understand which accounts to hunt which accounts to farm what is the propensity of that specific category across different demographics and all of that data help them significantly improve their uh uh the revenue in that category uh one specific point I'll make is the rate of reordering that the customers do you know has increased significantly because of a lot of ml based models that we run and we integrate that insight into the apps that the sales speak people use in their day-to-day lives and
• 29:00 - 29:30 that just that in itself has significantly increased their reorder rates so that's essentially paid for whatever they had to pay you so they got a very instant Roi yes yeah you start Us in the first one year I will show you the ROI on your investment so scan thank you so much for spending some time with me I really appreciate your insights it's a fascinating story and obviously you guys are going to be tremendously successful very soon thanks thanks and thanks for that great conversation and
• 29:30 - 30:00 yeah uh looking forward to you know more such and sharing our uh growth story with you that's great thanks all right [Music]