The AI-Ready Data Team with Erik Duffield

Speaker A: Early on, when AI first started popping out in this, um, generative AI space, a great diagram came out of just describing as like the fifth generation language abstraction. And I really think that that paradigm resonated with me. Like we've been abstracting complexity away from software engineering forever. This is just another way of doing it that makes it more accessible to more people, but they're still masters of it. Um, and so I think it's, it's a great opportunity, uh, to be in the data space. And if you're a softw engineer with data experience, you're in a great market.

Speaker B: Welcome to the Data T, our monthly podcast on the news, trends and technology transforming data today. We're your hosts, Armand Petrossian and Satish Jayanthi, the co founders of Coalesce. Let's get started. It's your first time hearing that music, isn't it, Eric? Isn't it great?

Speaker A: Outstanding.

Speaker B: You gotta love it. You gotta love it. It's one of the things I look forward to. Uh, but I've been really looking forward to this conversation. It's a complete pleasure to have you on. I can't believe this is your first time on the Day to T podcast. I would love if you could quickly just do an introduction on yourself and your role and then there's going to be a lot more questions I have about your founding story and everything there. But if you could just do introduction for yourself and then we'll get rolling.

Speaker C: You bet.

Speaker A: It's, uh, it's fun to be on here with you guys. Yes, correct. We've had a, a friendship and a working partnership between the companies for a long time that's been just outstanding, genuine and successful. Um, and so maybe this is the uncut or the cable version, uh, of the podcast that needed, uh, clearance to happen, but really enjoy it. So glad to be here. Eric Duffield, um, one of the co founders, CEO of Hakoda, now Hakoda and IBM Company. Um, I've been in the data space for about 30 years, so I've seen three major generations of technology evolve with different companies, um, uh, over that time. And it's been a lot of fun, um, and that journey, really unexpectedly, I did not know that in my career, another generation of technology that would really start to shift things would come up and create the opportunity for Hakoda. But I'm glad it did because it's been an amazing time to do this with partners like you and Snowflake and the modern data stack, um, and just a lot of fun to work with. Clients on everything they've got going on right now.

Speaker B: Awesome. I'm excited to. I got a lot of questions for you. If you don't know who I'm at, who I am. I'm Armo Petrosian, CEO, co founder of Coalesce. I got my, my co founder here, Satish. I'll let you introduce yourself real quickly and we'll get rolling.

Speaker C: Hey guys, this is Satish Jayanthi, uh, CTO co founder, Coalesce. Super excited to have you here, Eric. Great to be here, Satish.

Speaker B: I've got to say, this relationship between Coalesce and Hakoda goes back to pretty much almost immediately after we came out of stealth. Hakoda had just launched into the Snowflake ecosystem. We had launched exclusively into the Snowflake ecosystem as a transformation partner. And I remember like it was yesterday meeting up with your other co founders as well as Imran for dinner at Re Invent and we were talking through the vision of each other's companies, the potential, where our focus was. And I can't say I've ever felt more aligned with a partner. And that probably is why you all have been our reigning champ partner of the year the past couple years when it comes to implementations not just with Snowflake, but also Coalesce. So couldn't be happier to have you on Eric, first thing for the audience, uh, you've got a great background and founding story with Hakoda. Could you share for everybody how you even got into this? I know you mentioned you've been in data in a long time, but what was the trigger point for you to go and launch into the Snowflake ecosystem and do what you did at Hakoda? Like what did that, how did that all come to inception?

Speaker A: Yeah, a little quick history, uh, that'll inform kind of my decision making at the time. So early on I was part of a data consultancy that focused on that first generation of technology, really pre the emergence of cloud. It goes back, um, so we were doing standard data warehousing and analytics for companies as those systems were really starting to get a hold inside large organizations. And then, um, that company grew and eventually sold right at the beginning of, um, the Hadoop, uh, time frame and distributed file systems. And I moved on to uh, an organization called Appirio, that was Enterprise SaaS. And it was really at that emergence of organizations adopting the cloud when most people were concerned about security. This isn't real, it's a fad. And we went through that entire cycle, uh, with amazing organizations like salesforce and Google and Workday, um, and saw enterprises adopt the cloud. Um, that company sold in 2016 and I went on to Deloitte, um, on the software side of Deloitte to create a solution um, using real time interaction management and customer data. And I had some really Sharp, there were three acquisitions, some really sharp CTOs that we together realized there's got to be a new way to do this. They by the time they were acquired been built on data platforms of different vintages and we, and we thought this can't look today like it did before. And along that we found Snowflake, um, and this was early on. So this is you know, early on in our uh, cycle with Snowflake. Snowflake had been around it for years. So this is 2019, early 2019. Um, we wanted to move to that platform and did. And I saw the vision that Snowflake had, um, not only for the technology, there's lots of great technology comes and goes. But when you see that combined with a market opportunity, a vision to expand their addressable market with new things like transactions and applications and you see an organization that's really mature and built for growth and what Frank at the time was, was building with the team was pretty amazing. So all of that in all of my history convinced me hey there is a company to be built around this ecosystem. Um, and so that's what we did. And it wouldn't have happened had I not gone to people I'd work with between 10 and 20 years in this space across different companies much like you guys that had the trust playbook that we could work shorthand and work very quickly to do something differentiated. And that's what we did. And I brough my co founders, Ryan, Bob, Patrick, Kim, all, all that, you know, um, and then we expanded that team to really the heart of our team is that group that, that came after that, the leadership team of a coda. And what I love about that story is you know there's lots of we're engineering based company and I really felt that this market needs engineers that really think about things and basically innovate their way out of repeatable patterns. And engineers bring lots of technology to us and say hey this is interesting, maybe I'll start using this. Um, but that's there's a difference between great technology and when organizations are aligned in their vision about what the market needs and how do we work with customers and what we're going to bring to customers. And that's really when Patrick, one of the co founders um, started seeing and investigating and working with you guys when he came back, like there's something different. Yes, there's great tech, uh, that's a minimum required, but there's organizational alignment around how we see what customers need. Uh, and that really aligned to our philosophy around Snowflake, which was predominantly making complex things simple. Lots of people can admire complexity and have lots of complexity, lots of knobs to turn, but making it simple is an art form and that's really what we thought the customers needed. We built our organization around that. You guys were very aligned to that and it created this uh, great go to market relationship where we could serve customers with a similar philosophy.

Speaker B: Uh, I love it. I would say all the credit as far as our product philosophy goes should go to Satish. But one of our core values, one of our six, is to keep it simple. And we've always been really passionate about taking complexities away but still providing extensibility. I saw a ton of similarities to the way Hakoda approached the market and really taking what usually is such a painful, difficult, risky, time consuming implementation when it comes to data engineering use cases and simplifying it for your customers, obviously great, mission accomplished. Hakoda was acquired by IBM. Now you've got a way bigger vantage point to so many other organizations. And I don't want to lose sight of the topic of this, uh, podcast. It's all around the AI ready data team. And AI is a topic, uh, of conversation in literally everything these days. Given you've got so many organizations that are working with you, you've got so much visibility into what companies are trying to accomplish when it comes to being AI ready. How would you even begin to describe what uh, what, what an AI ready successful data program looks like?

Speaker A: Uh, I think it starts with people recognizing what the shift to AI is because some of it existed prior in the machine learning phase. But really it's a shift from processes that are making data consumable by humans to making data or the majority of data consumable by machines. And that's a vastly different um, process. And one of the things I mentioned about re platforming a product, uh, when I was with my prior firm to even work we're doing re platforming products now on Snowflake is that ratio of how much of this data is going to be consumed by humans versus the machine. And it's not just lifting and shifting to new technology. The entire design pattern is different in order to do that successfully. So I think that's the biggest thing to be AI ready is to understand We've got to think about prepping data, serving data, governing data that a machine will consume and that a human will participate in. And so now if you said two years ago 80% of data was consumed by humans, large uh, data warehouses, semantic models, massive bi sprawl. That's because we're trying to feed humans in all these different ways. Now it's starting to shift and there's still a long way to go, but the technology's there that we can feed machines to be integral part of that decision making. I think AI ready means you've got to understand fundamentally that how's my architecture going to change, how's my organization going to change when humans and machines have this orchestrated experience where it's a combination of using data together. Part of that is clients that are successful, companies are successful, start to do some research to understand what are humans really good at around context and using analogs. What are machines really good at? Calculations, um, consistency. You have to think about it in that regard and then understanding that you understand what your outcomes are, your goals or your endpoints are and you start to build the processes to for the machine. So it's the data governance, it's your pipelines, it's your auto healing, your monitoring, all these things that the systems are now really good at that are very different architecturally design pattern miles away than what you would have done five years ago had you started to build a data system. And so that's really where organizations start to determine their long term success or failure about being AI ready. Yes, the shortcut is we've got to have all our data ready and accessible for machines. But there's a larger kind of thought process around it that I've noticed in organizations that are truly successful. They're really thinking about that, um, those elements of it. How's our team going to work? How are our consumers going to use the data? Where, what decisions are we going to turn over to machines? What decisions are humans going to be involved in?

Speaker B: That that makes sense. I agree with you. I'm curious, you mentioned we've still got a long way to go to close that gap. What gaps do you see between hype and what data teams actually deal with in reality right now? If anything comes to mind immediately as they go through this journey and do they have too high of expectations when they get started? Like do customers have this kind of North Star that's realistic. Talk to us a little bit about that.

Speaker A: Yeah, I think if you look at kind of an AI adoption curve, um, or an educational curve to pick yours out there. But go to your favorite AI tool, I'm a huge user, and look at that curve and you'll, you'll see there probably is expectations early on for the past year and a half, two years, have been a little high, um, not in the long term, but in the short term, um, of what's going to be accomplished, how transformation is going to be in the short term, because they underestimate how hard it's going to be and the shifts beyond just the technology. So once again, we're in a technology adoption pattern where the capabilities of the technology outpace the readiness of the organization to adopt them and all the things it means beyond lines of code. Um, and so that creates this expectation internally in organizations and with customers, that this is going to be super fast, super easy, and it takes some practice to get there. So organizations will get there, but it means sometimes reskilling, shifting talent around, bringing in new talent, new ways, ways of working, new ways of running projects, new, um, new skills in house that are needed to kind of execute these programs. So I would agree it's been a situation where expectations have been really high. We've gone through the period where we've experimented a lot, finding out what works, what doesn't, what's harder, what's easier, not just in technology, but for us as organizations. And I think we're getting, we're probably pretty close to moving past that, to really starting to see major impacts. And I know everyone's enamored by the MIT study recently that says 95% of programs fail. Um, yeah, that's not our experience. Um, and so there I could see it when you broaden the set to everything that's been an experiment, um, because people are trying things with no intention of bringing it to production. And to be successful, they're learning, they're experimenting to learn how is this going to work, what are the boundaries of AI, where can we use it, where can we not? And that's all a natural part of the lifecycle. And as we work through that, to this production life of AI, where it's making real decisions, um, in real time, I think we're, we're getting ready for a very exciting phase that new skills are being developed, new patterns being deployed, new technologies adopted, that more of those promises will come. Now, uh, in the next phase, they're going to feel a little late because everyone wanted them right away. Um, uh, but there's material advantages created by being an agile organization that can deploy, uh, these capabilities over and over again.

Speaker B: Yeah, it's kind of reminiscent of how the data market evolved in itself. Right. It wasn't until cloud data warehousing made it very accessible for all types of companies and the supporting platforms and tooling to allow data projects to be successful. You could argue similar level of sentiment to how AI is evolving, where the tech was a little bit ahead of the skills and labor and the way companies were ready to operate. And so we're now seeing that same maturity curve over the course of what will be the next year or two. And you being on the forefront of that makes complete sense. You have visibility into that. And so I'm confident people will scale up, uh, teams will evolve. But it does take time and that's something that hopefully moves as fast as possible. Uh, so let's say that's the case. As you're talking to data leaders in the industry, customers and AI is rapidly evolving. What do you think that they lose sleepover today? If it's not the potential functionality of tech like, uh, what are they? Where is the pressure coming from deploying AI projects successfully since the tech's there?

Speaker A: Yeah, I think it's still um, number one, if you're really talking about losing sleep, data, uh, security. Um, right. So think about the speed at which data is moving in a machine or AI generated. If your pipeline and your entire process flow, your data ops is AI automated, it's moving very fast. Um, and so data security is always the big one. That's the job killer, the company killer that everyone has to get comfortable with early on. I'd say the next thing that's probably more surprising is the composition of their teams. So teams have been built and hired and developed and nurtured over last generation technology and organizations. When they get you into the quiet, senior executives, when they get you into the quiet room by themselves, that's the question they want to know is do I have the right team to execute this? If not, how do we shape it? How do I use part of your team? How do I reskill, uh, my existing team? What skills do I need to add? That's always one of the early conversations because it's key. The senior leaders know it's key to project success and that's really the marker that everyone has to put on the board, the points on the board later. Um, at a macro level, for us as humans, as governments, as organizations, I think the labor force shift is the biggest concern. With what's going on, how is labor going to shift? And I would say when I think through that and have some really fun Conversations on it. The part that always gets to that's kind of meaty that we kind of, we could address is if AI uh starts to replace the junior ranks early on, how do we develop and create the senior ranks? Whether it's a software engineer that no longer. Right. Is not growing through testing in a QA role into an engineer role, into a senior engineer role, or it's an attorney that starts as an intern, then an associate, then a junior partner and Right. So it's that those are the things that we're quickly automating and scaling. Even if It's a, ah, loop 10xing it. There's one human and nine used to be humans. That's at a macro level what I think people are thinking through. But at an organizational level, is it secure, is it governed? Um, and can my team really build the modern. Do they know the modern patterns and how to launch it? And so we're not just taking what we know and applying it to AI?

Speaker B: Yeah, yeah, that makes sense. It is interesting. The entry level positions or what AI is targeting, whether that's in IT or sales or you name it. So how do you ever get to a senior position if you never had a chance to be in the entry level position? Um, I'm always optimistic, otherwise I wouldn't have started a company. But I'm sure this will get figured out. I don't know exactly how or what it's going to look like, but uh, that is something that's been on my mind as well. Satish, I know I've been asking all the questions. I'm sure you got some. Otherwise I've got plenty. But anything you're curious about for Eric?

Speaker C: Yeah, yeah. Um. Uh, I know you talked about there's a lot of new kind of new ways of using the AI uh technology. You're seeing some success there. Any special kind of use cases where people having success versus some use cases where they're, you know, they, they thought it would work but they, it just didn't. Like for example, we hear a lot about like AI agents, you know, where things are being done in an autonomous fashion. Right. You rely on the AI uh agent to do certain actions. Are you seeing any of those type of types of use cases in production? Uh, or we say we are still in the experimental phase for that?

Speaker A: No, in production, definitely. I think, um, the agency use case, the success or failure is not whether it's agencic or not. Um, it is what it's being applied to and how ready is that? Uh, so the core Technology to create agents or systems of agents that um, process workflows and perform work. It's happening now. There's a rapid conversion to this right now. You'll see it mainly on the cost takeout side. That's the early phase because it's controllable. Um, then it'll move to the innovation side, the revenue generation side. But the ability to create the technology to create agents successfully, to monitor agents, to train, to um, to adjust them, to retrain them is all there. I mean companies are using it and there are successful case studies and there are ones that aren't. Part of it is the, did they have, did the company or the organization deploying it have the right goal in mind? Did they have the right context in mind? Do they have the data prepared? Is the data cleaned and governed and ready for it? But agents among themselves is just a new paradigm for processing, um, and scheduling and pushing some of the decision making into the system. Um, we're seeing it be very successful. It's not a determinant, um, for success. What you choose to have the agent do more is. And we're seeing cool experiments in IT that don't work out, but you learn a lot and you point it at something else. And that's really that early phase of boundary testing. Where is, where can I apply this or not? I think it's very meaningful. I don't think it's a loss to uh, for organizations to experiment and not move something to production if they learn from it. Um, but there has to be that organizational realization that it's uh, is when you're experimenting, you're learning from your failures as you are from your successes and taking that and informing what you do next. And I think there's a lot of that happening despite the um, the recent headlines of where people are thinking it's less successful. There's no backing down from it. Everyone's learning a lot and they're reapplying the technologies now with some skills, um, in the areas that it's very suited for.

Speaker B: Yeah, yeah, it's, it's one of the coolest things has been seeing how our customers leverage AI from their actual data implementation. So there's a couple I have that come to mind. But even simple examples of making support phone calls easier for customers at an insurance company, tons of AI use cases there to accelerate and improve wait times, responses, response times, everything.

Speaker A: So yeah, I mean call deflection for larger organizations saves them tens, hundreds of millions a year. Um, yeah. And it's a better customer experience quite Honestly like when you, everybody, something fell down if you weren't able to self serve or serve within an interface. And so I think there's great uh, steps being made in those examples.

Speaker C: Absolutely. Yeah.

Speaker B: Cool. So we talked about some of the concerns there around roles for entry level positions uh, potentially. But how do you see the roles of data professionals changing over time? Do you have a strong stance on that? Strong opinion on how you see the, the role shifting.

Speaker A: It's a great outlook for, for data professionals as long as you're prepared to go with the flow. And so the um, what I would point out on this interesting kind of um, early career path and what do we do with early career is the, let's take engineers that Satish works a lot with day to day. Yeah they are adopting AI faster than anyone else. And so right. The senior engineers, they've got a lot on their plate. They're managerial, they're architectural, like they're, they're not getting the opportunity to be, to be as hands on in the deployment and working with it. And so those fresh engineers coming out of school have great technical skills. They're applying it on their own very, very fast as a, like a digital native, um, using that technology. So my expectation is they will grow with it. I think your entry level attorneys will be using AI as there is their left and right hand. It will be very natural to them. And so the maturity of each industry uh, in its adoption of AI depends on having that digital native, that AI native worker, deploying it and running it with them and matriculating through the ring. So I'm very optimistic around that. But we have to find our way there and recognize that there are some big labor shifts coming. I think for data engineers this is just going to make data more important. Data is moving faster, it's more impactful, it's getting more um, more decisions allocated to it. And so the people that shepherd it, that design it, that architect it, that test it, the Q, uh, the validate it, um, it's great careers for them. And so the early on when AI first started popping out in this um, generative AI space, you know a great diagram came out of just describing as like the fifth generation language abstraction. And I really think that that paradigm resonated with me like we've been abstracting complexity away from software engineering forever. This is just another way of doing it that makes it more accessible to more people, but there's still masters of it. Um, and so I, I think it's, it's a great opportunity uh, to be in the data space. And if you're a software engineer with data experience, you're in a great market.

Speaker B: Yeah, yeah, it's like, ah, I would imagine of all the roles that could be negatively impacted, a data engineer is the lowest on that list. Like it's such a, uh, I mean it's such a important skill right now. And sure they may be able to automate or accelerate a lot of their work, but they're still going to need to always be a highly specialized data practitioner at companies to help make sure that it's delivered with high quality, trust it, the logic's refined, it's working properly. I just think that that would be the hardest thing to replace as this continues.

Speaker C: So.

Speaker B: Yeah, can you agree more? Yeah, exactly. Um, what new skills or mindset do you think data leaders should prioritize then in that case?

Speaker A: Um, I think, well, one, I think data governance is interesting to think about and talk about, especially in the context of being AI ready. Because for a large part of my career data governance was a control mechanism and now it's really an enablement mechanism. Um, it was locked down. You can't access this data. This data has got to die at a certain time. Like it was all about control and control and access were at odds and now control and access are aligned. You can give more access and have more control. So you got to think about that and plan for that. Now what are you going to do with that and you do. In an AI age where data is moving through a machine and it's not going from human that manages the pipeline, what happened? Okay, I need to reload this data that is going away. If you design your patterns well, to the BI analyst who's updating the semantic model, to the BI analyst who's writing reports, to the frontline managers who are consuming reports or pivoting data around all a lot of that work's going to flow through machine and then it's going to grind data a lot further for the human to interact with it and make decisions around analogs and around context that humans are great at. Ah. And so I think organizations need to think about and people need to think about governance as an enabler. And what is an enabling data governance process look like along our entire data value chain?

Speaker B: Yeah, governance has had a bad rap for a very long time. Like I think data teams have been, they've been resistant to data governance teams like data engineering teams. It's like the last thing they want to do is put in more bureaucracy and red tape. Whereas, uh, that's Been historically data governance's team's roles is to try to slow things down and put in figurative laws into how the data team is supposed to work. Whereas now obviously we went through the acquisition of a data cataloging company and are seeing how data governance teams want to move now is so different. Like they want it to be an enabler, not something prohibitive. And I expect that trend to continue as organizations have to figure out how to do things with data in a way that's governed but without slowing down the operations. And it's been a, we call it governance shifting left. And I think that that's uh, a trend we're going to see continue to grow a lot more quickly just as a result of AI initiatives.

Speaker A: Shifting left's uh, a great way to visualize it because even if you think of data layers, bronze, silver, gold, gold is always that ah, walled protected garden that uh, no, it can't get to gold layer unless very it governed, restricted, hard to get in. Which means the CIO's organization generally has very protected trusted data and marketing goes off and does their own and creates their own because they've got to move fast. Right. And so now you think about machines. Is, is are your algorithms only going to access excess gold data uh, that's incredibly curated along that process? Likely not. So now you got to think of governance in this world. Well they're gold was our curation process on the human consumption path. Now we've got to think about layers. And there was a really good question in the Q and A that someone posted around. Do we need this middle layer? Are we sourcing data right from source systems? This is that exact thought process of. In the AI world. Yes. Uh, just like in the machine learning world. For those organizations pre AI that were uh, pre gen AI that were spending a lot of time on machine learning, you're accessing data from lots of places, not the final curated uh, data source. And so um, shifting left on governance and really enabling data to be accessible by humans and machines very quickly, very trusted is. That's a great role to be in. And one talking about great career paths, truly being on data governance, um, is, is a very strategic role inside organizations that I think is also a cool path.

Speaker B: It is, yeah. Satish. When it comes to trusted data. People use this term all the time. Anything that like, like how do you even define trusted data? Because you know, Eric's talking a lot about like executives and making them making decisions, data being trusted or governed properly. Like how would you, how would you define that Looks in An organization. Kira.

Speaker A: Yeah.

Speaker C: I think there's two steps to get to the trust. Uh, one thing in my experience, it has always been the consistency. So first of all you got to make the data consistent. Whether it's right or wrong. It has to be the same reported across the company. And then you start talking about whether that's right or wrong. So there's a two step process in my opinion, because if it is inconsistent, it just gets very, very hard even to like, where do we start? Um, to get that consistency? The first step. But then obviously the definitions, uh, you know, the data governance comes into play here is what we mean by these things. Uh, and the second part of that is where are we getting this information from? Like what sources, what data elements, what calculations, you know, uh, who owns this? There's a lot of things that come into play, but um, yeah, that's what my experience has been. Trust is built over time, uh, first with consistency and then um, with proper definitions and proper, uh, transparency.

Speaker B: That's definitely a way to measure success, I would say, right? Like there are so many ways that trust can be eroded if you do have inconsistencies. How would you measure the success of an AI ready data program, Eric? When you go to customers, like, is there a general rule of thumb around what it looks like when it comes to speed, trust, business adoption, cost efficiency, how would you define an AI ready data program being successful? At the end of the day, um,

Speaker A: maybe it goes beyond AI ready. But ultimately I think the measurement for the overall data program is the marginal time cost, uh, to the next interaction, to the next AI thing that you launch. Because at a program level, it's not, not just the one thing. If you're on an AI journey, you're creating multiple agents, multiple interactions, you're pricing, your pricing is being handled by agents and AI. Your um, shipments are being handled by AI. Right? So it's, it's endless. And so if, I think, if you're, if you grade yourself, am I A.I. ready? Uh, do I have a modern data system? How fast can I create one and move to the next one to move to the next one to move to the next one? One.

Speaker C: Right.

Speaker A: And so each one of those along the way, the aggregate of those need to be successful in reducing cost or driving new revenue. But from a platform basis, from a team basis, that marginal time cost to deploy the next thing, I think super important because if, if that's really low, you're adopting new design patterns you've done in data with date, your data ops. What software did with software DevOps over the years, right? You've integrated the system, you've automated it, you're monitoring it, you're adjusting it, self, uh, healing, all of these things. And I think that should be the goal. If, if everything is put into just producing one thing, your first AI product or service, um, you won't recognize the importance of the system that you're building and building the next one. And that's the most valuable thing as an organization. If you're really good at deploying it and you can deploy 10 in the time it takes someone to deploy one, how many do you need to be successful to just crush your competition? You could have 3, 5, 6.

Speaker B: It comes back to what you were saying earlier too. Like 90, uh, 5% of AI projects failing. It's like, so what if the intention wasn't to succeed on the initial one and you can change rapidly and have 10 experiments in a shorter period of time and one of them is a huge hit, then that's really what you need to be looking for. It's around how quickly you can iterate, how quickly you can adapt and find success, even if there are certain failures along the way. I know internally we did a ton of experiments with AI, uh, and then we found some incredible use cases that it was very relevant for. But you don't just choose the right one right off the bat. And if you did, I would be skeptical because there's probably a lot more gold to uncover in the path anyways. And so, uh, yeah, I think the adaptability is something that's highly relevant for data teams and any team look at innovation.

Speaker A: Lowering the cost of failure. Right. So if you can, you can do those 10 experiments faster and cheaper than everyone's doing one, you've lowered your cost of failure. You're, you're, you're now an effective experimenting organization. If you're experimenting organizations, you take six months to get something approved, you go through all procurement, everyone chokes the living life out of the thing because they're paranoid of failure. You're going to drive failure and you're never going to get a new one. Um, and so I think that's the pattern we've seen in other technologies that in, in successful organization and we're early in this phase, that's real. What really impresses me is that lower cost of experimentation and failure, um, allows you to drive more successes.

Speaker B: Yeah, yeah, that makes sense. I agree. I think it's something that organizations either have an abundance mindset towards or a fearful scarcity mindset towards. And it Perpetuates itself, doesn't it? Like you see, as you start to find success, it starts to snowball. Whereas if you are paranoid and constantly afraid of taking a bet or you're close minded when it comes to being experimental, that's when things dry up.

Speaker C: That's right, yeah. So it's never been, it's never been easier to do that though. I mean, I was just going to say I remember anything that we had to try in the past used to take several months. So that's the fear of experimentation. You don't know the result, but the experiment itself takes so long that you're just not scalable. But yeah, absolutely agreed that this is the right time to take that kind of approach.

Speaker B: I'm assuming this is back in the day when people were drawing logical models before building out the physical data engineering use case and ah, a production instance would take six months to go live, if not longer. But obviously as cloud adoption evolved and modern technologies evolved, that became so much faster. So I expect a very similar renaissance when it comes to AI. Uh, speaking of predictions, we got a couple questions left and by the way, anybody in the audience, if you've got anything you're curious about for Eric, feel free to hit the Q and A. But I also can't believe that it's this time again. So every year we do data predictions for the upcoming year. It feels like the year flew by because I'm back to being here asking you, Eric, what are your predictions when it comes to significant data trends in 2026, which I guess only three months away now that we're at the end of September? Anything that you're looking forward to or maybe just trends in general that are a top of mind for you as we turn the corner into 2026?

Speaker A: There's a few that are particularly interesting to me. Some will have longer term impacts, but I would not be surprised if you think about how hot and frothy this space has been if there's a little bit of a shakeout in, in calendar 26 in the valuations of AI.

Speaker B: Oh yeah.

Speaker A: I mean it's just been so hot. How can it not be? And of course it's.

Speaker B: Wait, you mean there's a bubble?

Speaker A: Uh, aren't there always? And bubbles are great. I love bubbles. Bubbles are fantastic, they're exciting. And I bring it up to say, look, if there is a shakeout, that's not a bad sign. I mean, obviously if you're concentrated in those, it'll hurt for a little bit, but it's a Natural part of the cycle of the market. Right. Things bubble up, they get super inflated, they ease off a little bit. But that's, that means we're moving from one phase to the next. And what we see with our clients is the amount of um, so take it from away from the AI companies and the corporate companies. The amount of efficiencies and productivities gained with AI are real and they're starting to show up and we're early on in that. And so while we may go through this public market cycle of AI software companies and foundational model companies, all of that, that may ease off a little bit in 26. Because at some point the tea kettle's got to blow some m. Steam, um, the, the client side, the actual implementation of it. Right. We're moving in that next phase of that cycle that I think super exciting. Budgets will increase, spending will increase, successes will happen, our clients will have better earnings calls, Their, Their uh, stocks will go up. Like that's a phase that I think it feels like 26 could show us some of that. Which uh, um. Never be, never be afraid to ride through that. If you're going to be a bubble guy, you got to be a, a, a post bubble guy. Let's call it that.

Speaker C: Yeah.

Speaker A: Um, I like, um, so now being part of IBM, I'm very quantum aware and I, I love the boldness that IBM put out around quantum computing and, and demonstrating next year fault tolerant, quantum. You know, with a release of, of this kind of in that 20, 29, 2030 time frame that's coming balanced. So that's, you know, we're three years away and there's going to be proof points, um, demonstrated this year by IBM and maybe others to follow. You know, thinking of AI and data readiness, you know, quantum, quantum AI. I love that the engineers are progressing and pushing forward in inventing things from the ground up, from the silicon up. Um, and So I think 26, with all the capital invested, with all the venture and private equity money and corporate capital, look for broader uh, technology innovation. You know, AI that is in the wake of this AI that feeds and kind of accelerates beyond that. So quantum is something I'd always keep an eye on. We're certainly watching it closely. It's, it's being experimented with some really interesting case studies. You know, we've got a few years where uh, broad adopt before broad adoption comes. But I'm super excited about that

Speaker C: for

Speaker A: the next bubble too. Don't.

Speaker B: That's what I was about to say. I'm more excited about that than I am about the bubble to pop. But I find it funny because both Satish and I, we just did a roadshow called Coalesce Join, where we had Joe Reese, uh, on a panel, as well as Joe Floyd, who you've met, one of our board members, but a general partner at Emergence, where they're making tons of AI investments right now. And the question came up, is there a bubble? And Joe Floyd, the Emergence investor, was like, there's absolutely a bubble. We just don't know what inning we're in. Are we in the second inning or the ninth? Is it about to pop tomorrow or are we just getting started? Nobody can answer that question. But who knows? Maybe 20, 26 will be the year it does burst. Let's see.

Speaker A: Well, whenever it happens, it'll rebuild. Um, this is not exactly not going away. Um, people having. Getting your data ready for machines to be part of the consumption path is AI readiness. And that, no doubt, is only going to accelerate for sure.

Speaker B: We got time for one last question. Eric, this is always a fun one. And I say that to you specifically because you are a CEO that I've always respected a ton, greatly admired, and your ability to paint a vision, follow through on it and execute. I think Kokota has done an exceptional job at doing that compared to not just consultancies that we've worked with, but any business period. Is there any advice or closing words for anybody tuning in, whether it's data related or even career related?

Speaker A: Um, as I think through this,

Speaker C: we're

Speaker A: in a world, and I mentioned up front, Dakota was built around engineers and how the impact that engineers will have, uh, in the marketplace for the last four years, in the next 10 years, um, now saying that if you're an engineer or you're a finance or business person, my advice is it's a great time to focus on strategy, um, focus on effective leadership, building teams, um, and then also brand, which you could simplify and think of storytelling, your ability to get, to communicate your message out there. So in a world where engineering is everything and there are, for the right people, $240 million signing bonuses, um, that are really good AI engineers, um, that's an incredible opportunity for a resurgence in the discipline around strategy, effective leadership and building teams, um, because that's what's needed to pull it all together. And so whether you're an engineer, a lead engineer, an aspiring leader, or a business leader, a sales leader, um, those skills are incredibly important to bring to reality everything that we know is possible over the next decade. And so I always like to say, okay, what are we not paying attention to? And if everything's about the bits and lights and the engineers, then that means there's got to be an incredible opportunity for those who have skill, uh, and discipline around strategy in leading people.

Speaker B: Right? Yeah, that makes complete sense. I love it. It's been such a pleasure having you on, Eric. It's always been amazing working with you and your entire team. I'm, uh, glad we were able to make this happen, and I can't believe it took us this long. But we're here now, and I just want to say thank you and congrats to all the success and the future success for you and the Coda team.

Speaker A: Thank you, guys. Really, uh, have enjoyed the partnership and will continue to do so. You guys are two of the best in the business. Um, always look forward to getting together.

Speaker B: Appreciate it. Thanks, everybody.

Speaker C: Thanks, Eric.