The Digimasters Podcast - Unpacking AI in Data Analytics: An In-Depth Chat with Keboola's CEO Pavel Doležal

Join us at the upcoming Empower event in London on 28th Feb 2024. Register today

Welcome to a riveting episode of The Digimasters Podcast, where we delve deep into the world of AI and data analytics with Pavel Doležal, CEO and co-founder of Keboola and CEO of CxO Lab – Rob Saltrese. This episode is a treasure trove of insights for anyone navigating the complex landscape of data operations, automation, and the integration of AI into business processes.

🔍 In This Episode:

• Discover Keboola: Understand the essence of a data operations platform and how Keboola is revolutionizing the way businesses handle data.
• Expert Insights: Pavel Doležal shares his journey, from the inception of Keboola to its pivotal role in data automation and operations, offering a glimpse into the future of data analytics.
• Empowering Businesses: Learn how Keboola enables companies to streamline their data processes, ensuring a human touch in AI's role within digital transformation.
• Culture & Growth: Dive into the culture at Keboola, the challenges of scaling, and how the company maintains its core values and spirit amidst rapid growth.
• Empower Series: Get a sneak peek into Keboola's Empower series, aiming to empower users with the tools and knowledge to become data heroes in their organizations.

📅 Event Spotlight: Join us at the upcoming Empower event in London, where you can experience first-hand the capabilities of Keboola, engage with industry peers, and explore the latest in AI and data analytics.

🌐 Stay Connected: Don't miss out on the insights from Adam Nagus, your host, and his discussions on how Digimasters utilizes Keboola for automation and AI to produce impactful content. For more information, check out our website and follow us on our social media platforms.

#DigimastersPodcast #DataAnalytics #AI #Keboola #TechnologyInsights

Transcript

Adam and Rob: 0:08

Hello everybody. And welcome to another episode of the Digimasters podcast. this week we're actually talking to CXO Lab again, and Keboola. the reason that I thought Keboola would be a great company to work with and to bring onto the podcast was because right now, when I've been building a lot of different types of automation, and, data analytics platforms, I've realized that. The advice I've been given about which sort of architecture, which products, which tools has become quite confusing because the market just has a plethora of different choices. Now that could be a very good thing, but it also means. When I was totaling up and, I'll introduce my guests in a second, when I was totaling up the total cost of ownership to put my dream platform in place, when I looked at all the different licensing and some were compute, some were storage, some were dependent on the number of cores you had on the cloud, it became quite an astronomical operational cost. And I felt that there needed to be something out there that could consolidate all these things together. So through my research and introductions, via CXO Lab, I met Pavel, the CEO and co founder of Keboola. Now, what I hopefully we're going to find out today is can Keboola as a data operations platform, or find out even what a data operations platform is. Be part of, or the complete solution for my dilemma around how many actual tools I need to automate business processes, data operations, data cleansing, and integrate AI in a safe way, into my marketing campaigns, my sales processes, et cetera. So let's find out. Thank you for bearing with me on a little longer introduction. As I said earlier, we have CXO Lab still with us. and I'll introduce Rob again, Rob Saltrese the co founder and CEO of CXO Lab. And Rob, do you want to give a brief introduction of what CXO Lab is again? I can for those who aren't listening for the second time round, in which case, apologies for repetition, but, yeah, we're a consultancy business, particularly really advisory that, help empower founders,even like really achieve and reach their North star, right? We utilize our fractional C level experts to delivered very specific bespoke tailored solutions. That helped companies accelerate rapidly and hopefully avoid some of the speed bumps and pitfalls along the way. Perfect. Thanks, Rob. And Pavel, for the delay, but, let's get around to introducing our main guest, our VIP. And that's Pavel, the also co founder and CEO of Keboola. Welcome to the show. Let's jump straight into it, Keboola, what is Keboola?

Pavel: 3:11

Hey Adam and Rob, thanks for inviting me. I really appreciate your time. So let's first start with what is Keboola are you interested in how did the name start? Or what is Keboola I will start with what we do and then the story about the name. Maybe if we have more time. So as you said, like you, I've been building a couple of businesses before Keboola we with Friends, built actually one of the largest portal networks back in the day. Remember Yahoo and Style, Microsoft Network. We did that across Europe. Then after that I built two other businesses and everywhere. All my businesses are really relying on data and ML to automate processes. Everywhere. It was so hard to get, the data. And I always had to have, dozens of engineers to actually connect all the systems and get me the insights. And if I wanted to do it myself, even though I'm the technical, they would not allow me, and when I met with Peter, my co-founder.

Adam and Rob: 4:13

Mm. Mm.

Pavel: 4:17

Data to cloud for companies like Red Bull for example. And they were, very good backend engineers and they, like some of their clients tried to ask them for data solutions, what we would call today to build a data platform. And they were very lazy and it was not their core business. So they created this script, bunch of script that actually automated that. And I was working for, through one of my angel investments I was working for a large bank in Europe. And the CIO asked me, Hey, I would like you to, not only bring me the cloud, show me how better it is, how better people actually work together.

adam-nagus_1_01-22-2024_160107: 4:54

I have these projects with hundreds of people, different locations and I don't dunno if they're profitable or not. So what I want you to do, I want to connect the data from Oracle database, zebra Google Calendars, and couple of, Like Excel files and tell me.

Adam and Rob: 5:10

So had,

adam-nagus_1_01-22-2024_160107: 5:11

lucky

Adam and Rob: 5:12

a stakeholder that had realized that cloud isn't just about making it cheaper. Cause I think we've all found out that cloud isn't necessarily cheaper than on premise, A complex use case that was value based instead of just how much can you, save me from my infrastructure costs? Oh, amazing. Okay.

Pavel: 5:49

He just laughed me out of his office, he was like, you're so full of, like you come here and say that the cloud way is the better way. And then you can do a proposal like you would be I bm right? Or Accenture for that matter, Adam. And, Give you three months and$3,000.

Adam and Rob: 6:08

And 3,

adam-nagus_1_01-22-2024_160107: 6:14

Like, that has to be something. Cloud has been around for eight years or whatever, and

Adam and Rob: 6:20

to integrate all those different platforms. And you said, okay.

Pavel: 6:26

I like challenge, but I, what can I say? I was, back then, I was not literally doing anything because I sold my last business and I've been just investing, like in, as an angel and helping my startups, and I was like, Hey, finally something, challenging and, hey, I know this, I've done this, so many times before, right? And I was like, Hey, time times have changed. So I went to my Silicon Valley network, I searched through all databases. I went to Israel, I just couldn't find anything. And I got frustrated and somebody recommended me, to speak to Peter. Now my co-founder. And I was like, Hey, yeah, but they are just an agency. They don't, he was like, yeah, speak to him. He's literally smart. He might know this, right? So I invited him into our office. And, and there he said, and I was like, Hey, I have this problem, like this bank has this problem and they need to connect to Oracle. Hey, no problem. I was like okay. And then I was like, Hey, but also they need to connect to Google Calendar through API said, yeah, that's not the, I was like, shoot, I've seen this movie before. So nothing is a problem. And and but I was literally out of options, so I just like, yeah, signed an NDA and I shared the accesses, to him because like I, I had a, like instance of that. And then, six hours later I was browsing through the internet. It was like around midnight and I checked my emails and there was a link and,

Adam and Rob: 7:58

So,

Pavel: 8:23

If people actually are together, are they more productive on, on the project, which is similar to this project or not, right?

Adam and Rob: 8:29

was from Oracle

Pavel: 8:30

And I was like,

Adam and Rob: 8:32

combined.

Pavel: 8:33

yeah, that was from Oracle. That was from Google Calendar. And some CSEs. Yes.

Adam and Rob: 8:37

you'd back of this, obviously back in 2015, you'd brought all this information together and already produced dashboards, et cetera.

Pavel: 8:45

Okay. And I was like, how did you do it? And I was like I'm using this visualization tool, good data. I'm like, I don't care about visualization. How did you put it all together? I was like, yeah, we have this internal tool

Adam and Rob: 8:58

be using,

Pavel: 8:59

that we've built. I was like, shoot, I need to understand more. So next morning, 9:00 AM I actually came to their office. I sat actually under the stairs like these ones in my home and, I never left.

Adam and Rob: 9:12

going on the phone, but I'll be in the chat. We'll

Pavel: 9:15

We were really, trying to understand where is the future going to go, like what is gonna be done? And we came up with this hypothesis that we saw that, a, we are in one of those unbundling cycles, right? You've been around it long enough, so it's like bundling, unbundling, right? There's an S Curve of adoption. And, so we saw that all systems thanks to driving force of cloud, all systems like Oracle, like systems are actually being unbundled by cloud systems like Snowflake. And there is gonna be a lot of new things. So we came up with, like all the, all the business processes are gonna be algorithm driven. And we're like, Hey, this has already happened in stock trading. This has happened in advertising in RTB and more and more. The cloud is gonna be penetrating normal businesses. It's gonna be happening everywhere, but.

Adam and Rob: 10:13

I'm doing diligence work, especially focusing on the digital diligence for mergers and acquisitions and doing a review of somebody's digital transformation, they have digitized as in they've got Oracle, they've got SAP, they've got Dynamics, et cetera, but they are creating data because they're going obviously through an interface and adding new customers and things like that. But there always is a person moving the data, or there's an ETL, extract, transform, load process, moving data around. But very rarely do you see the actual business process completely converted and automated. So that's, is that the gap you saw, which was a case of, there are digital systems out there. People move data into data warehouses and then people move data into reports and Excel and do lots of manual entry. But the whole culture hasn't changed to true automation. So you can actually realize the benefit of what was written in that business case for that digital transformation. That is that's the gap I'm hearing.

Pavel: 11:23

That's one of the gaps. Exactly. There's a third part, and that's the growth of the data being created with all the SaaS applications and now ai. So if you combine these three forces together, it's a perfect storm. Shoot, what's next. And in order to actually really you, you put it very nicely in all of those digitalization programs that were written by big companies, consulting, like they know what to do. It's quite straightforward what to do. It's not a magic, but how to do it, that is very hard. And like it's very similar situation nowadays, because.

Adam and Rob: 12:05

I'm going to

Pavel: 12:08

they would've old mainframes sometimes, then they would have, like eighties stack then they would've nineties data warehouses, then they would have data lakes, and Hadoop, and then around that all of the modern data stack and streaming. And some part would have Databricks, some Snowflake, some would have, Salesforce and Salesforce would say that they do everything, but it doesn't really connect here. And I want to, just create my customer journey to understand who comes to my retail shop and who, clicks on my advertising and how can I target them, in personalized thing, like personalized. And I have to call my DevOps, like we call, admins DevOps now, because we still have to call them. So DevOps works a little bit differently in data than it works in actual, software engineering. And I have to wait.

Adam and Rob: 12:58

Rob, from your experience, obviously. Looking at the skills over the last, when data science quite a very popular role, what we did with data scientists and, I put my hands up and I think all my peers should put their hands up and say, including the recruitment industry, we did it. Badly, as in all of a sudden, these people coming out of university with data science degrees, we stuck them in labs, inverted commas and said, play with the data. And they played with the data and we got the business case all these, graduates and very big salaries because everybody's very excited about data science and the revenue or new products could bring in or the operational savings they could bring in. But we didn't really realize that because nobody knew how to operationalize what they did. Yes. They built this great Python routine or statistical IT when what's this. So from my perspective, what Powell's talking about is they came at this the other way around, instead of getting all the clever people together to do the data science and the magic, they said, if people are doing the magic, how are they actually going to realize that? How is the automation going to work? And if it doesn't understand Python because they're all. net, how's it going to work, and if you need data and need to be clean, so actually realize how's it going to work. I'm saying I'm putting words in your mouth, Pavel, but that's what I'm hearing. you were talking about they were hiring these young data scientists to be the magicians. They weren't magicians. They hadn't done it before, nor did they have the tricks they'd learned already in order to go and be a magician. Now, if we take this back into this world, we're talking about the data Pavel was talking about, all these different people, all these different siloed pieces of information, but manual people stuck in the middle, like a, a conductor operating different levels of train tracks and moving people around. What Pavel had was the information and a problem and a real case study to go and fix, which people clearly hadn't looked at before, and then was able to with your colleague, actually build something that was enabler. Really what you are doing is enable people and organizations to understand what data they truly have, where is it, how to access it. When you've got that. you can create everything. And then the kind of magic happens. If I look back in 20 25 years ago, when I was looking at this field, it was all around ETL. That was everything that we were helping companies hire for. This was big banks, insurance companies. They're all looking at You hiring plumbers. Exactly. So we were hiring people. Yeah, but absolutely, they were putting together the data blocks to segment it into certain places and then made it accessible, but then you couldn't move it, and that kind of was the point. Then you got the BI piece, then that phase came afterwards, and then that was the interesting phase in their mind around look at the reporting and that we can present you, but it was only on a certain layer and quantity of data and quality of data. That was available, but that felt like the next layer. Then we had the data science phase. And as you say, rightly. No, they've already done it. So what did they do? They hired people on 45, 55 K straight. And the rest. Yes. To be fair, I do remember telling, some younger people and people, university get into this space because it's going to be one of the next big things. And it is a, it's a good earning potential. It's a great place to be go. Cause data will be key. It's going to be the next oil. It's going to be the next gold. Sure enough. Pavel then has seen it through that point where it really started to. I think hit the meteoric rise button probably at the same time if you were doing what you're doing was when blue prism and all the other big rpa companies started to really take off and receive their obscene amounts of funding to do relatively tasks but without necessarily again the enablement to have the access to the data

Pavel: 17:07

Yeah, there's so much to

Adam and Rob: 17:10

I was just so much information you gave us. It's just, it seems so obvious with hindsight. And from my perspective, back in 2015, I was still one of these people that were given the problems by banks or, telcos, et cetera, and you were looking at I need a, an informatica to move all this data around. Then I might need a, a messaging system and EBS or. I said the right way. Yes. B I can't remember which round, enterprise serial bus enterprise bus. Yes. So let's

Pavel: 17:41

enterprise.

Adam and Rob: 17:42

or Tibco. and then I needed a data warehouse and it might've been still, an on premise behemoth, Teradata. That was a perfect one. Yes. I did a lot of Teradata at and then, we were then moving things into the cloud, but it was Redshift and then eventually Google BigQuery. the visualization tools. Is it Qlik is it Tableau, et cetera. And, yeah. And then we've had the streaming to come in, like Kinesis and obviously different types of Lambda architecture, but there were so many different components, as I said, at the beginning of this episode and now. If I'm, I still do quite a bit of solution architecture. That's why I'm having this conversation. So I use Keboola to do my ELT or ETL, moving the data around. I can actually do some data catalog work within Keboola. I can push it to. I'm hopefully not any but I'm hearing a marketing CDP engine, get the information back. Then I can actually send it to chat GPT to do some work and then bring it all back. And I can actually produce data that's for the business to use in their own self service. So if they wanted to go to any of these new RPA tools themselves. You can use Keboola to make sure that this is safe data sets. We've already got rid of all the PI data and confidential information. and so you're sitting yourself right in the heart of the architecture. Does that make sense? Is that right?

Pavel: 19:15

Yeah, that makes perfect sense. And I think we can now spend hours and hours to what Rob was saying, to what you are saying, Adam. Yeah. It's about integrating things together, putting like a layer, abstract layer on top of existing infrastructure because as we identified right in the enterprises, you have layers and layers of old architecture, which doesn't work together. So what we did, we created an operating system, if you will, like you have in your, PC that kind of operates different parts of the infrastructure inside a PC without user actually having to connect directly to the hard drive or allocate memory. And we cannot bring this experience to data stacks or data platforms. And, that means that you can actually add different kind of infrastructure and the thousand, the processes still run. But what it also means is that, the integration of that together is important. But what is important also is governance. As you rightly pointed out, like once you prepare this, you need to give it in a, governed workspace on one click through, let's say active director access to marketing department or personalization department as a data product, right? So they can, but not only the data set, but the whole product. All the ETLs, all the orchestrations, all the machine learnings, like everything, right? And they need to be able to click, have one workspace where They can, modify it. They use it. a little bit. Still govern, so security knows exactly what is going on and then reuse it

Adam and Rob: 20:56

My internal CISO,

Pavel: 20:58

on the work.

Adam and Rob: 20:59

Officer brain's going, okay, so we're expecting more regulation out around AI. we've got to prove that the model is transparent. So anybody that's wanting to do this, they need the full lineage. I'm also going to be a bit controversial if you're a CIO, CTO, or even a CFO that's approved all the spend on these other tools. You can still save face because it's not a case of ripping out and breaking all these deals and people going, why did you spend all that money on the infrastructure? You are a layer that. Can be brought in under different guises, all valid ones, such as security, such as optimization, such as transparency, such as being able to not just move data, but actually execute workflows. So that's, that's a message I would give to my peers saying that yes, you might have these contracts in place. You don't have to rip out snowflake. It's not a new database. You don't have to rip out snowplow. using, Streaming. This all works and you're the, semantic layer. as you said, I like that analogy operating system, holding it all together and coordinating it and putting that transparency lens on top of the existing architecture. And if you're a startup or scale up and you don't have all this legacy, then you can start straight with Kabula. Okay, perfect.

Pavel: 22:22

It it's exactly

Adam and Rob: 22:25

I was just thinking about, my journey, focusing on visual analytics and running different visualization, companies, adoption. Was either really easy, really hard because adoption was, Oh, it's very shiny. And we love this drag and drop. And we've been looking for, an app feeling around our insights. But sometimes the UI and UX was very specific to a person or a team. So some people loved it. A lot of people could hate It And it was very hard to bring in that sort of design from, let's say Apple, that is supposed to be intuitive to as many people as possible when you're actually building a dashboard or an analytics app and. What you're talking about makes a lot of sense to me because I'm thinking about you're not trying to be the shiny front end and beauty's in the eye's beholder. You're practical. You're, do you want this type of data? Do you want it cleansed? Do you want that extra insight added from an LLM, a chat GPT, and you want to know it's safe and you can go and play with Oh, and, who in the company doesn't want that? So your adoption should be very high. But my question is, who was your sponsor in the bank to get that sort of level of adoption so quickly?

Pavel: 23:46

Sponsorship. actually in in, in the Checkers bank, our sponsor was the, we were working under the CIO and the person who was responsible for data tribe platforms. Like that's that. And always honestly, our experiences, when you want to change the company, really change. Not just have another tool. You need to have a C level, like alignment. And not only oh, let's do it, but shoot, I am so frustrated it hasn't been done for last five years now I want data to be sitting under me, like with the board and actually working with the board. And I be, it's not only about technology, it's about the change

Adam and Rob: 24:30

one of the things I want to speak is not just yet, but why did you decide to go in as pretty much what I would call a low code or no code platform? So when using Keboola, I'm building these routines and I don't need to learn any codes. So it's very intuitive. It's almost like you aimed it at the business instead of technology people. but before we get onto that, I think Rob and I were talking earlier and we're quite interested to know if there were any issues or challenges, that Kabula went through on its journey from advising companies, building an internal product, and then realizing. Ooh, maybe the internal product is the product. So how did you make that decision? And has anything that you could share gone wrong and any lessons learned, especially for other people out there that are on their startup scale up journey?

Pavel: 25:28

Okay. Thanks. Thanks. That's a very good question. When I joined guys, 2015 Keboola was consulting agency, like for movements to cloud. you have to understand the consulting business was pretty sizable. It was, I don't know, million dollars back then or something like that. Not bad, it was growing by its own, but there's, through my engine investments prior to this, I've seen that there isn't always, a problem between, comp in the companies that want to do both.

Adam and Rob: 26:01

I'm going

Pavel: 26:02

Because like dynamics of service business are totally different than dynamics of product or platform business. And, I believe starting a product from consultancy is a great way to start because you understand the problems, right? You are like you know why you did this because you are so fed up with, right? So you have a burning problem. Second, if you can connect it to the, strategical framework and form a hypothesis where it's gonna go,

Adam and Rob: 26:38

You've got to think about, are you just creating a product? Cause you like it and hopefully building it and somebody will come and buy it. But because you were a consultancy, you knew the problems you knew was out there. You used, you put yourself in your customers, existing customer shoes, and that's how defined what, Keboola's product was going

Adam N: 27:01

to be

adam-nagus_1_01-22-2024_160107: 27:02

yeah. was like, we started to see that, if you grow the consultancy business, it's very hard to say no to that money. because it's so addictive, right? It's just every month you gonna get this done. But,

Adam and Rob: 27:39

because you just look at the utilization and the chargeable. Can you charge this person out? And you obviously make a profit on the person. A product business is it's all up front costs. It's huge risk. And was that a

adam-nagus_1_01-22-2024_160107: 27:50

Yeah.

Adam and Rob: 27:52

from, changing the business from people to a solution and from a mindset from you as founders and then the employees, because that is, and it wasn't the planned one, was it? So you got to an inflection point to say, Hey guys, where are we going?

Pavel: 28:07

When I joined, when I started to work with the guys for a year and something I really wanted to understand if it has a potential scale, I didn't want to do another service business. Been there, done that,

Adam and Rob: 28:18

Bye.

Pavel: 28:19

it's great, but no, I just didn't want to do. so in like year and a half looks like, Hey guys, we need to decide, right? It's a great consultancy business. If you're gonna use this tool as your internal tool, you're gonna go and grow to 10, 15 million in consultancy. It's gonna be great, right? Just not of an interest to me, because I believe that what we have here has a potential to actually change the world and change how people actually work the data, because all of these, let's do digital first. You know, It ain't gonna happen.

Adam and Rob: 28:51

talking

Pavel: 28:52

People in business are not gonna be involved and we have a potential to change this. And so the actually the decision was easy. It was surprisingly easy. My co-partner, my partners are very smart. So like they saw it, what was hard

Adam and Rob: 29:08

Transcription Outsourcing,

Pavel: 29:16

service, right? And we said, okay, let's do this. And for the first year and a half, after we look back, at the year, we're like, so we got into 90% of service and 10% of software. That's not moving the needle. Like everybody wanted it, but to change, like practically, it was hard. So we had to go the other way. And we like really hard. We just pushed the services out. we gave them to our partners, and we said, okay, now we're gonna bet all in on the product.

Adam and Rob: 29:50

business

Pavel: 29:51

we went from 90%.

Adam and Rob: 29:53

trying to convert the services business, you did a trial run, you saw things out there, you were still very much focused on services and then you span it out and split it. And Keboola went the

Pavel: 30:04

Yeah. And we, I can go in details, but there was a lot of service around Yeah, exactly. Building the overall implementations of data strategies and dashboards and everything, like helping companies. So we said okay, so we had two, three partners and we said, okay, like we want you to do this. We'll give you this business. And let's talk what the license for Keboola should look like. And we want to focus, we focus on the product so we can better service you. And it was very painful for the first year and a half. And, we wanted to do it on our own. We didn't want to take any external investment because two things. First, we needed to go through this transition. And second, we knew that we are a year or two years maybe ahead of the market. Because you remember back then, 2018, 19, 20, everybody was like, oh, just have this one tool and it's gonna sell this, bring this one tool to the already, like very complex thing and it's just gonna solve this one thing and it's gonna be perfect. And so we knew we had to wait until this was

Adam and Rob: 31:08

unbearable to

Pavel: 31:11

people are like, shoot, now I have 30 different tools. My business users can't, really use it. It became so spaghetti and my C level is asking me, so where is the transition? And so it thanks, not thanks, but due, due to covid and the digitalization, like really speeding up in Covid.

Adam and Rob: 31:32

do

Pavel: 31:33

and started to, like really nicely take off and we started to go bigger and bigger enterprises. So we took, a first round of financing actually in 2022, the angel investment. And and it was from our network of actually people who used Keboola to build their businesses like Thomas Trooper and Raleigh Group. It's a almost$2 billion business all around from Keboola in four countries. They in, in Europe, they deliver food, like 20, 20,000 SKUs within two hours, and they have zero people for data platform operations. They don't have them, they don't have, data DevOps people, they don't have, platform people. They have Keboola it's a automation of, how you run data platforms and literally operating system. The same situation as was in the, we talked about this before, as was in the analogy, in like when the telephone came out, people had these huge switchboards, right? They were done manually. You had to call their switchboard. There was some young female usually, which would switch it manually. And then, at after 10 or 20 years, there was 250,000 plus minus women actually working at switchboards in US. And there was news articles, that if everybody's gonna adopt telephones, there's not gonna be enough women to work in the switchboards as operators. And it's just kinda like similar situation to. To data stacks now you have to have, you have to call DevOps and they have to manually put things together and it was suddenly changed via At&t, In the switchboard industry. They automated and now we don't have to have somebody sitting and switching, two cables together.

Adam and Rob: 33:23

I might not be old enough to completely remember in the 1920s, but, yeah, the way I it is, the bottleneck you talked about earlier. When people in business are saying, why is this so hard? Why can't we do it? And I'm just thinking about the data is that bottleneck And as a chief data officer, we are always seeing, why can't you just give us the data or why isn't the data ready? And we go back and say it's because you're giving it to us in a very poor state because you're creating sales data and customer data in a terrible way. And IT's problem. And then IT says, no, it's the business problem because everything is people in there and the more people touching it it. the downstream impact is it gets slower and there's just not enough people they're doing. but if it was automated end to end, there was business processes were written down and the, the solutions, the software, the digital platforms were all. Checking all the data, only letting you input things and removing human error as much as possible, then us in the data teams be giving you, data sets for you to enable all these exciting processes. So yes, I'm going okay. The the original telephone operators and how they were plugging in all the different jacks. And then when that got automated, that was when Telco can actually scale. Thank you for that. From a perspective of growth in the times that we've talked about with COVID and everything happening and kind of people being locked in. But. How did you, and I go back to that word, which relates to something you're going to talk about later on. So enabling and empowering both your internal culture and then the adopters of the technology that you were selling to your customers for them to use it too. How did that work?

Pavel: 35:08

When we started, like around 16, 17, you like really focus more on the product. We need to hire more developers, and more people to work with us and stuff. And, we set couple of, like truth for us, We said we don't want to turn into another Oracle and, in, not in the size like they are awesome, right? But in a way, we said, look, we want to couple of principles. A we want to be, for our clients always to be, have easy migration path, right? So one of the things that we embedded into the architecture of the system is an API where you can actually take out the data with just one API call all the data, all the transformations, everything you can just Take somewhere, right?

Adam and Rob: 35:58

As in you, you don't proprietary data structure, what you put into Kabula. It's very easy to take out. So you're not locked in. Got it.

Pavel: 36:05

Yes. Yeah. because we as a, running tech companies or tech divisions, we hated that always with our vendors, right?

Adam and Rob: 36:14

to

Pavel: 36:15

Second, was like, like as much as possible, code should be open source on the course, not on the core, but on the outside kind, right? If you look in GitHub slash Keboola, over 90% of our code is actually open sourced.

Adam and Rob: 36:31

And

Pavel: 36:32

of our, culture was like, no BS. we also wanted to have kinda like whole personalities. People who are not only work, but also have personal lives because like we believe it's important. If you are only work, then you're gonna burn out

Adam and Rob: 36:46

easy to say. So

Pavel: 36:48

So we were,

Adam and Rob: 36:49

within Keboola to make that easier.

Pavel: 36:51

when we started, we were looking for people who have their hobbies. So a lot of people were, for example, mountain bikers, but like extreme mountain bikers. They would move to our office in Vancouver, so they could be mountain bunk biking. So they were some of the best software engineers ever because they had a passion for, what we do, but they also wanted to desperately, get out, in the wild and do some biking. I've been doing these 360, right? And actually my 360 is available for anyone to comment on, and give me feedback. Oh yeah. I get very harsh feedback. Yeah. And I love it, it gives me an opportunity, to, like actually go and reach out to the person and, Hey, what did you mean by this?

Adam and Rob: 37:37

It is, you almost look for when, we've talked to you about this and Adam and I talked about it in our podcast the other week around a fractional career and people with multiple things that they do, multiple hobbies, they enjoy going and doing many different things that often allows them to bring far more sort of intensity and capacity to their actual work. So rather than just focusing on work and thinking that is the best way of doing it. Actually, it's the complete opposite these days and really you want to do that. Plus you're giving them the freedom to have an opinion, to grow. And that's the right thing, right? That is culture in itself, right? And culture is no one thing. It's the people you employ and what you empower them to do. And if they can all feel like they've got a voice and be heard and have an impact, which let's be honest, everyone wants to do, particularly the new generation coming through now, it's their biggest thing. Is I want my voice to be listened to and I'm bright, which is why you've hired me. And I've said, I don't think this is a great idea. What can we do about it? Here's a suggestion. And those are probably the that you want to do. And, it's interesting. You mentioned it. I actually had a, I had my own 360 only a couple of weeks ago with Nico, my, my business partner. And again, it was, actually a very rewarding thing to do. It was really, Help crystallize a few things. It brought a few things to the fore, really allowed you to understand where all of your true kind of powers are and actually how to keep building upon them, which is good for the business, and good for your own career, but also the areas where you may have. Gaps or things you may not even spot that you can then actually tweak and tune to yeah, it's

adam-nagus_1_01-22-2024_160107: 39:08

We have a Nick side, right?

Adam and Rob: 39:09

He French, but he lived in Holland. He's about to say Having it. Yeah, having that having a Dutch person give you 360 feed back. It's quite an experience. it's a double whammy as he jokes French mentality with the Dutch mentality mixed together. So again Very honest. Are you okay? I'm fine. It a quite, exhilarating and liberating thing, like genuinely. And I went away from it and Making notes and looking at things and it was good for both of us I gave him the same thing in return and actually we both left Within that kind of last week going right these are it's a great way to do things and pavel But they're still not talking to each other

Pavel: 39:48

It's only been two weeks. It's

Adam and Rob: 39:50

It's four now. It's developing it's a break. We call it a rest period And, and how do you see that sort of transitioning? I know this is something talk about. obviously now launching the, you did very well, huge, great, fantastic round, which was, phenomenal. We'll let you talk about that. That's now allowing you to, develop in, into the UK and open the London office. like another challenge, in a good way is how to do it, how to build it successfully, and then how to try and keep that. mantra and culture going without necessarily being there day to day. So how do you see that panning out?

Pavel: 40:26

It's kind our third iter or layer, like where we need to reinvent ourselves again, right? As you grow the company, you need to kind, keep reinventing yourself. what we are the most worried is kind like how not to lose the spirit of Keboola in, in the process. And,

Adam and Rob: 40:45

Outsourcing,

Pavel: 40:46

in next 12 months. And so it's very hard. So here are a couple of our things that we've been, really focusing on and I would love to know the comments, like maybe in comment sections, people might have experiences, how to do it better. couple of things. So first

Adam and Rob: 41:03

to use the

Pavel: 41:03

we, year and a half ago, we really,

Adam and Rob: 41:07

want

Pavel: 41:08

structured our hiring process.

Adam and Rob: 41:10

So, I'm

Pavel: 41:12

we wanted to first go, like with top grading methodology, but top grading is very hard. And I haven't, I talk, talked to a lot of people, it's so hard, but people don't actually do it. It's one of those things that it sounds very nice, like when you logically perfect,

Adam and Rob: 41:32

find that companies say they're doing agile and then no, you're not. And you're forcing agile methodology on a process that really shouldn't be agile, but it's very cool. Yes, Sorry carry on

adam-nagus_1_01-22-2024_160107: 41:45

Sorry we actually learned that the son of the guy who wrote, top Grading is doing Who? Who, and, it's simplified version of top grading.

Pavel: 41:57

the process is actually has six, six stages. And the last one, like in between there's a meet with the team. The last one is who interview, which takes three hours. And, we really invested into this. It's very intense you have to involve two people full-time throughout the whole process. And then you have a team, of people that needs to spend hour to two hours with the candidate. So every candidate you spend with them, 10 hours maybe, like throughout the process. And but Looks like for us, right? And which people will struggle if they join us. And we went, we went, we actually decreased the, number of non fitting hires. Like we decrease that to something like below 10% and that's really good, but now we need to scale it, right? So we need to keep the system second how we can like, did it. We've been working with Oli our CRO. For year and half now, first year we worked together on consulting basis, so we got to know each other. We really made sure that it's not only about skills will, but it's also about, the culture, like dimension. And when we started to hire for London, Ollie is like a center figure, right? He understands, he knows, we are very well synced together, how to do things, and we can build around him. Third point is, we, with the first batch of first 20 people, we actually spend a lot of time together. So we bring them a lot to check. We had an offsite, in, in, in late November in mountains in Czech where we brought, over a hundred people and all the new people that were just about to join. our of sites are a little bit different than most companies of sites.'cause most companies are, in my experience, they mistake off site for company conferences. So company conferences is the brainwashing, right? But literally the content of the offsite is done by team. So we have on Confluence page, we have everybody who wants to talk about something. Everybody who wants to run a workshop, it's like unconference, right? Because like we, when we did the first one two years ago our people team from US back then was like, how can you do that? We need to put together the agenda. I'm like, Hey guys, we spent so much time trying to hire the best people ever. And then we don't even trust them to put, the agenda for offsite because like we, we think we know better than them what they need to talk about, right? So it's not, it's never gonna work, right? So it works perfectly. Or everybody who wants to talk about topic pick, list, topic in confluence, or describes the outcomes, describes, who they would Then

Adam and Rob: 44:53

The,

Pavel: 44:55

topics that get voted, actually get a time and space and people can sign up for them. And then actually they have to present the results of the workshop and it needs to be done.

Adam and Rob: 45:06

the word empower. So I wonder if you could tell us a bit about this series of event called empower and why the UK second event.

Pavel: 45:25

and not only hackathons, those are great, but they are just for some part of the user groups, right? And by the way, it goes to your, question Adam with no code. And, but we don't have only no code. You can actually use CLI, and you can do everything in part of your ci cd pipelines because it's about collaboration between tech people and business people. But that's like a topic. And we set, like our users actually asked to do kind meet with us, see what's new, do some workshops, how they can implement things, talk to other customers, right? And we're like, Hey, how should we call it? And like we, we did and we asked around the company. And, Palo, a guy from product he came with, Hey, we are all about empowering people, empowering normal people with just common sense in the business and some technical skills to be data heroes. That's what we do every day. You take, person who has common skills and some little bit of knowledge. We teach them a couple of SQL queries and within three days they do wonders. So that's how we should call it. We should call our series of events for our clients, with our clients Empower, because we are literally about empowering people. We decided to actually do London Empower as well. We did a practice run in Prague and it's gonna be in February. Everybody's invited. I think it's gonna be very pleasant experience, how to actually, talk to other people from the industry, talk to our clients. There's gonna be a lot of clients showing use cases. Also see, firsthand what is capable about, maybe actually do something on site, come out, with some problems being solved, being empowered. And yes, there's gonna be some sneak peek into our AI initiatives as well. There is a, Adam, you said it nicely, like in the, like when you come in more and more with the AI things and you need more governance, full lineage. And we've been working on something, how to use LLMs not only as a

Adam and Rob: 47:30

Oh,

Pavel: 47:39

them to understand the. Line and, governance of everything that is inside Keboola. So you can then ask them who build this data set? What is an impact if I change the SQL code here? What is this metric and where is it being used? So all of that we are gonna be showing there and I think it's gonna be awesome. Of it.

Adam and Rob: 48:02

NLP, natural language will be part of Keboola so you can ask Keboola about the lineage, as you said, and it'll respond to you in natural language instead of having to hunt through log

Pavel: 48:15

Yep.

Adam and Rob: 48:16

Really? Very clever. I haven't seen that yet. Anybody else do that? that makes a lot of sense. So you're going to be building, obviously, LLMs based on the data by Keboola, at your, obviously, different client sites. Interesting.

adam-nagus_1_01-22-2024_160107: 48:34

Have to remember as an operating system, we have all the metadata, right? Because they go through our API layer. So we are able to actually use LLMs to create a knowledge graph of what is going on. And then you can query it in natural language. And it is, yeah it's really nice. Oh, sure. Yeah.

Pavel: 48:55

We did a sneak peek, a couple months ago that we are working on that. and, I hope that we're gonna be actually able to demonstrate that on some production, projects in London.

Adam and Rob: 49:06

on the 28th of February. 2024, just in case anybody's listening to this podcast in the year 2032 and gets excited. and, yeah, But you never know with a podcast or a YouTube video, when somebody watches it and they go, Oh, that sounds great. up. And, you realize that's from 2019. Yeah, I'll put a link, a registration link in the description of this podcast and also and I'm you look for Keboola LinkedIn page and website, you'll be able to find it as well, but I'll tag it in the bottom of this episode.

Pavel: 49:42

it. I was just like in my head when the scenario when you would be listening in sixties to Minsky where he was talking about the neurons, he called them differently. And you like, oh, this is so we should do something.

Adam and Rob: 49:53

I think we've got so many different stories and we've been at least recording for, I think an hour and a half at the moment. So it's going to be a fun editing this down to an hour episode. And I don't know what, but, and we've got so many other notes that we could go about, but we've ticked off, obviously I understand company. Rob discussed with you about the culture of the company and then I'll see. The people that are listening to this, they want to understand how to do a startup and scale up. There's some interesting elements there, but obviously from a data operations perspective, I'm very excited because our other sister podcast, the Digimaster Shorts, which is the daily podcast that uses a lot of automation. In fact, it's 90 percent automated. And I'm using lots of different tools, et cetera, but the heart of it is automation platform. And I've started to rebuild some of it on Keboola because the limitations I had on tools like Zapier and make. com was they are good for small companies. But when I wanted to integrate Into my architecture for large enterprise or private equity with lots of data. They don't have the right connectors. They don't have that Support model for large transactions and business ready. It's fine for a startup and playing around But I can't actually implement it even with you know The CISO constraints And SOC 2 compliance and choosing which VPC I didn't have that option because it's only SaaS, a lot of them. but with Keboola you can bring it into your own VPC. It can be fully SaaS, et cetera. So that's why my network being able to scale it to medium, large enterprise and portfolio companies. It made a lot more sense for me to migrate off these purely SaaS. options onto Keboola. So hopefully I'll get to talk about that maybe at the event, cause I'm coming along and you've asked me to speak about something. So maybe I'll be talking about Digimaster shorts powered by Keboola. That'll be exciting.

Pavel: 52:00

That would be actually exciting because I, that you can then, take what you've built and turn it into a template and register template to our marketplace. There's over, like 400 developers that have registered something to our marketplace. So all of the components in people, most of them are by third

Adam and Rob: 52:18

I'll be creating competition for my own podcast by selling a template and how to generate podcasts. It's another

Pavel: 52:26

You, you're gonna be multiplying yourself.

Adam and Rob: 52:28

You just got to look at it with a different lens. It's a, how you can build through CXO Lab Digimasters. You can build your own AI powered podcast series. Yeah. I'm going to do an actual episode on the DigiMasters podcast, this podcast channel, because it's long form talking about how we built DigiMasters shorts and the decision we went through. And obviously that's going to be marketing and technology, but at the core of it, it is, how do you do daily podcasts? Cause there's a lot of effort that goes into it and you have to do it with automation, unless you've got. Zero else happening in your life. Otherwise you'll continuously, producing podcasts and marketing them doing nothing else. if you're going to do something daily that's my opinion anyway.

Pavel: 53:14

Adam you don't even know how big of a theme this is for me. because I've been backed, through my company for years that I should be publishing more. And, I understand like in this day and age it's part of the CEO job, right? I think, I don't like to,

Adam and Rob: 53:33

is going to be

Pavel: 53:34

I don't I'm not, like it's not, it doesn't come natural for, to me, I like theater or small audiences, but, doing something publicly on LinkedIn and, doesn't come naturally. But I also understand that every time I do some piece of content, which actually has some, audience that it's helpful for the company. So I actually was like, I need to automate this because Otherwise I can't, do it. So, that's another conversation.

Adam and Rob: 54:01

keep the human in the loop, mainly because the nerd in me likes to see what DALL-E is going to generate as a thumbnail. The creativity side, I'm going, how did it come up with that photo? And sometimes I'm like. That doesn't make any sense. And other times I'm like, that is genius. So it's always exciting, even though it's automated for me to see what's going on. And if I hadn't automated it and built it all myself, there'd be no excitement. So every morning I get to say, what has the robot the developed, and generate. To summarize the podcast episode I've created the day before that goes live every day at 6 a. m. So I get that excitement as well as the automation, even though I'm the producer of the podcast in the first place. So I'm looking forward to sharing lessons learned, around scaling gen AI and automation, even through, my own projects within Digimasters, let alone with my clients, et cetera. But we're going to have to finish this episode at some point, because I don't know about you, but, I'm getting very hungry. I think you're in the UK times at the moment, aren't you?

Pavel: 55:11

I'm in Spain in Canary Island Santa. I actually promised my kids I'm gonna take them for the Indian food tonight.

Adam and Rob: 55:17

and have Indian for dinner and I can't because I've got something incredibly healthy and I've also got to go for a run so these don't really quite equate to your eating. Can I not talk you into it Rob? Can I say? Nope. You don't want to nip out for a quick curry? No, this seems to be a theme. We do a podcast and I go for a run, so I'm going to stick to tradition and I'm going to go for a run.

Pavel: 55:35

But you can have some Indian food and then you can go for a run

Adam and Rob: 55:38

Pavel, when you're in the UK in a few weeks, we will go out and I'll, there's a should go to. So we'll, we'll do that and we can and we can meet up, but agreed. Yes, we probably should

Pavel: 55:49

but can we take

Adam and Rob: 55:50

definitely like to, in the follow up, we can talk about the event. We can go into a bit more detail and very much from the CXO Lab kind of community side of things. We'd love to delve into how you can help startups, utilize right. So they can take it in from a very early stage of an organization. Obviously that will talk to our very senior kind of fractional C's that we've got on our founders and the C level to go and look at it as a tool and a technology that they could implement too. So it'd be great to look at it from that side of things, but we can. We can delve into that one. I just want to come to the event to find out why you came up with the word Keboola and if it means anything in check and why a massive octopus, I get the tentacle thing, why blue and, on my little architecture diagram, saw I generated, with DALL-E a little octopus tile for an icon. Very cute, but we'll have to save that for another episode. I think

Pavel: 56:45

yeah it's cute.

Adam and Rob: 56:47

so anybody, if you've enjoyed this episode And you are in, You don't have to be in the UK, but if you want to find out about automation and AI and more about Keboola obviously you can listen to future episodes of the Digimasters podcast, or you can actually come in person and meet all of us, I think it's at the OXO Tower on South bank in London on the 28th of February, but you must register. And so link will be in the description or on the social media posts. So do register and do come along and meet us. Now, thank you again to both my guests. Oh, Rob doesn't feel like a guest anymore. He's like part of the furniture and, but definitely, but literally stuck in that chair now. Pavel and, we'll talk to you very soon, but thank you again for coming on the Digimasters podcast and we look forward to finding out tons more and speaking to you again soon. Hope you enjoyed yourself.

Pavel: 57:37

Thank you. Thanks for the invitation. It's been a real pleasure. I know we scheduled this for so much shorter time, but you guys have such a great questions that we could do this for hours. So thanks everybody and

Adam and Rob: 57:49

Bye bye. Bye bye.

Pavel: 57:50

Cheers.

Don: 57:50

Thank you for listening to The Digimasters Podcast. Be sure to also listen to our sister podcast channel'Digimasters Shorts' a daily podcast covering all the latest news in AI, Technology and business. remember to follow our podcasts and turn on notifications so you never miss an episode. You can follow us on instagram@DigimasterShorts or search for Digimasters on Linkedin and facebook, you can also email us at podcast@digimasters.co.uk