3491: From NHL Ice to Enterprise Data: Ataccama's CEO on Building AI That Actually Works

[00:00:00] Now, running a network that publishes more than 60 interviews a month does take a lot of work behind the mic, and our sponsor helps make that possible. So a huge thank you to them for believing in this mission and supporting the stories that I'm telling here every single day of the year. If you're building a startup, every second counts. And sadly, so does every breach.

[00:00:23] And the truth is, early stage companies are often the easiest targets because attackers know that their defenses could be thin. And NordLayer gives you the same level of network security that some of the big players use, without the premium price tag or complex setup. And best of all, it scales with your growth. You can add users in seconds, protect your assets from day one.

[00:00:46] And over the years, I've seen many founders lose investor trust over one day early. So it's just not worth the risk. So why not secure your startup properly right from the beginning? Check out NordLayer.com slash TechTalksDaily and get 28% off with the coupon TechDaily-28. But now let's get today's guest on.

[00:01:13] What does it really take to make AI work inside a business? Is it cutting edge models, massive GPU clusters? Or is it something far less glamorous, but far more essential? Well, today on TechTalksDaily, I'm catching up with someone who's been answering that question long before it became the industry's obsession. Today, I'm talking with Mike McKee, and he's the CEO of a company called Atacama.

[00:01:40] They are a global software company that helps enterprises clean, govern and trust their data. So that AI can actually deliver results in the real world. And believe it or not, it's been seven years since Mike last appeared on the show. And in that time, the world has completely changed. We've lived through a pandemic, witnessed the rise of generative AI, and seen data move from being a back office function to a boardroom priority. And Mike's journey is nothing short of remarkable.

[00:02:10] He began as a professional hockey player in the NHL before transitioning into tech leadership. But now, as the head of a company backed by Bain Capital, he's helping global brands like T-Mobile and Prudential build the clean data foundations that every AI system depends on.

[00:02:30] So today, I've invited Mike back onto the show to discuss why the companies succeeding with AI aren't the ones with the flashiest models, but the cleanest data. And we will talk about the real meaning of AI readiness, the cultural shift in how CEOs view data trust, and a few of his personal lessons on leadership, mentorship, and pragmatism.

[00:02:54] So it's an honest, practical look at how businesses can turn data chaos into clarity while creating AI systems that people can actually trust. So much to get through today. So enough from me. Let's get Mike back onto the show now. So a massive warm welcome back to the show. Seven years have somehow passed since our last conversation. And I'm not just saying this.

[00:03:21] It feels like a year, two years back since I remember talking to you. But can you tell everyone listening a little about who you are and what you do? Because I suspect we've had a few listeners on board since then. Absolutely, Neil. No, it's great to be connected again. And I remember my parents always saying that time starts to go quicker as you go older. So it's not a good sign that we're both staying seven years feels like a year or two. But it is the truth. And good to be back.

[00:03:50] I'm still in Boston and still working for a company with Bain being the primary investor. Seven years ago, it was Observant, which became part of Proofpoint. Now it's Atacama, which is a standalone, obviously, data management company in a fascinating area right now. You know, Bain invested before the, I call it the chat GPT moment in late 2022.

[00:04:15] And all of a sudden, boards and CEOs and executive teams started caring a whole lot about data and the quality of that data. And there's a lot more demand for what we're doing right now. And I think a lot more good can be done with it. And, of course, we have to mention your remarkable journey from playing professional hockey in the NHL to leading a global software company. So, again, for anybody that missed our previous conversation, how did that transition happen?

[00:04:42] And are there any lessons from the rink that maybe still influence how you lead today? Well, to be totally honest, Neil, the transition's still happening. I was trying to sort of play with the young folks and I racked up my shoulder a couple of weeks ago. Well, so I haven't given that up. And what turned out, I didn't know about coming in Atacama, but there's a lot of hockey players at Atacama. Obviously, lots coming from the Czech Republic.

[00:05:09] And the fact that way back when, Jarmo Jagr made me look bad, helped me ingratiate myself to the folks over in Prague. About half the company is over in Prague. And Dominic Kaschik's from there, too. I played against him. But I say it all the time. Software is a team sport, just like hockey is a team sport. Whether you're a goalie, a forward, a defenseman, the whole team has to come together. And it's exactly the same in software.

[00:05:35] So whether it's our development team in Prague that has to work with our go-to-market team around the world, from Australia to London to Toronto to Boston, it's not successful if we're not communicating openly with one another and working really well together. It's, like I said, just like hockey. And looking back when we last spoke, I mean, what, since we last spoke, we've had a global pandemic. We've seen people working from home at scale around the world.

[00:06:04] Open AI is coming up to three years since that dropped as well. So much has changed. But your mission is still rooted in data quality and governance. You've been shouting about this stuff before it was cool. But for listeners who might not live and breathe this kind of world every day, how would you explain why data trust is such a big deal right now, especially as we're entering this AI phase? Yeah, no, great question. Gosh, it is amazing that a pandemic occurred between the last time that we spoke. Yeah.

[00:06:33] And it's funny you use the word cool because I was at a CDO conference here in Cambridge, Massachusetts recently. And the CDOs of Vanguard, New York Life, JP Morgan, the US base were all there. And at the beginning of their panel, they were talking to each other saying, we're the cool kids now. I used to be included in meetings. It used to all be about regulatory compliance or defensive use cases.

[00:06:58] And then to your point, since OpenAI came into the forefront, people obviously have to worry about regulatory compliance. But when it comes to better analytics and it comes to using all the models with AI, you just have to have that data. So it is an area that I used to, I somewhat joked that it's gone from being CIO-driven data projects to CEO-driven data products.

[00:07:26] And all of a sudden, CEOs and business leaders and boards are paying attention to these projects. And for so long, going back further than seven years ago, people say, oh, I don't trust the data. I'm not sure about the data. That's not an acceptable answer anymore. You have to be able to trust that data if you're going to drive your business on data,

[00:07:48] drive it on analytics and obviously feed LLMs in a way that you can trust the output from those models and leverage AI to its full extent. And I think many organizations are saying now that they want to move faster with AI, but they're discovering that it's their own data that's actually holding them back. And I was at a conference last week, Jitex Global in Dubai, and one phrase I heard about three or four times is without data, there isn't AI. There is no AI at all.

[00:08:17] So what would you say are the most common issues that you see when it comes to messy or unstructured or unreliable enterprise data? What are the big things you see here? Yeah, great point, Neil. So I like that without data, there isn't AI. You know, I go back to the age old people process and technology and AI and Gen AI is like a new technology and it's great. It's incredibly powerful. We use it a lot internally. It's obviously in our products. It's core to what we do.

[00:08:48] But if you haven't thought through who's going to use it and what are the new processes and the governance around it, the technology is not going to be that useful. So in some respects, and it's also tricky with AI because its use can be so decentralized and sort of organic, which is what you want. But at the same time, you got to figure out how that fits into current processes that exist within companies. So in my mind, and it's funny, it's, you know, without data, there isn't AI.

[00:09:18] And I'm also a big believer in organizations like without organization structure, things don't happen. So who owns AI? Who owns the use of AI? How is it going to be measured? And that measurement piece is just so, so important. Like everybody will, you know, AI readiness. Let's go. Like, okay, what does that mean? You know, what areas are benefiting? What decisions are being made? Why are those decisions being made? Is it driving more revenue? Is it decreasing costs? Is it decreasing risk?

[00:09:46] Really coordinating between the data teams and the business teams and answering the why. And then thinking about the people process and technology when it comes to the how. And I think when companies talk about AI readiness, we always hear about GPUs or model training, but not enough about that data quality. So I've got to ask, why do you think this part of the conversation is still often being overlooked? What's going wrong here?

[00:10:14] Yeah, no, really good question. And, you know, well, because it's hard to a certain extent, it's hard to measure quality. Now, we have seven different dimensions that we measure in our software from uniqueness to completeness to freshness to accuracy to validity. So there's lots of different measures and you can tweak those measures to get to your data quality score. But you need different data quality scores for different sets of data.

[00:10:44] You know, your bank balance, you probably want to be exact. You know, a marketing campaign and address that goes out and your web surfing habits in different websites and spots, you know, that doesn't necessarily have to be as exact. So it's hard to figure out what's an acceptable quality for different sets of data. And it's easy to say we need better quality data. We need to be able to trust our data.

[00:11:11] But connecting those dots and prioritizing, OK, which business initiatives and which areas of the business are fundamentally making subpar decisions and could fundamentally run better with better data and really prioritizing all the data. It's funny, like your accent, I was with one of the biggest banks in the UK recently, and they look at all the data quality initiatives that they have open and they calculated how long it would take to complete them.

[00:11:41] And the answer was 2030. Wow. Completed. And like, that's not a good answer. So what you have to figure out is like, OK, from this sea of data quality issues that exist, which ones need to be prioritized and which ones are getting resolved within X timeframe?

[00:12:01] And really boiling it down to what are the most critical data elements that are out there and how do we measure whether those are getting fixed or not? So it's tricky, you know, going back to that people process technology, but thinking about all of those different things and not accepting the sort of like it just has to be AI ready or it just has to be higher quality. Really digging in once again to the how you get there.

[00:12:26] And I was reading that you've mentioned a few times that the company's succeeding with AI right now and not the ones building the freshest models, but the ones with the cleanest data. We've been talking here about some of the bad examples of what to look out for and why we're failing here. But are there any examples of what good clean data looks like in practice, that utopia kind of ideal there? What does that or what should that look like?

[00:12:53] Yeah, no, I mean, we have customers that literally will have a data quality score against, you know, tens and sometimes hundreds of programs. And they'll have a threshold for those different programs that the data quality has to hit. And it's funny, I'm thinking about one now, which is in the spirits distribution of all things. And two years ago, when we started dealing with them, there was this general feeling of distrust in the data.

[00:13:23] And once again, this was a great example where the data team worked very closely with the business leaders and said, hey, what data don't you trust? OK, let's run a quality score against that. And let's see how you feel when that quality score moves from 60 percent to 80 percent or 60 percent to 95 percent. And then after about a year, it got kind of quiet. And, you know, going back to sports, you know, referees are doing a good job when no one mentions them.

[00:13:51] Data teams are doing a good job when no one's, excuse my language, bitching about the data or the data because the inverse of that is they trust that data. So once again, very practically connecting the data team to the business, understanding what data set they care about, running the quality checks against that data. And then good things happen.

[00:14:14] And the same organization is also leveraging AI in what I think is a very real way, whereby, you know, salespeople will go into one area. That area normally buys ABC spirits and wines and people that buy ABC want DEF. And once again, that's a function of high quality data. And they're seeing much more cross sell. They're seeing much deeper customer relationships.

[00:14:40] And as I said, the sort of general myth or complaints around can't trust the data, very quiet right now. And looking at some of your client lists here, I mean, you've been working with some major players from T-Mobile and Prudential that particularly stand out. So just to bring to life what we're talking about here, how are you helping brands like this build AI ready data foundations?

[00:15:04] And what kind of automation or governance capabilities make that kind of possible, make that possible at scale? Yeah, no. And you mentioned the word scale. I mean, T-Mobile literally scans 22,000 databases and 17,000 applications every week. And, you know, they've got regulatory concerns in terms of PIA data that they need to be on the right side of. But, you know, they are so many. It's so funny.

[00:15:33] I mean, so many companies are essentially data companies and they live and die by retaining customers. So if they can get any information around behaviors that are predictors of losing a customer, it's incredibly valuable information to them. T-Mobile also is a massive consumer business and a massive commercial business.

[00:15:58] And not surprisingly, there's a bunch of people that use T-Mobile for their home phone plan that use T-Mobile in their business, yet they'd never really cross marketed to the two. So now you've got a single set of data and you can trust that data and you're scanning that data. You can make a lot better decisions. I'd say the same thing for Prudential or many other insurance or asset managers out there. I mean, there are literally insurance data going back over 100 years at a lot of these places.

[00:16:28] And this sounds kind of morbid, but sometimes there are people that are dead that are still in the system and they probably shouldn't be in the system. So cleaning up that data and making sure that there are master data records, that data is quality from an underwriting perspective, from a recoveries perspective, is just essential. I mean, once again, I keep saying insurance companies are effectively data companies.

[00:16:54] So to do their job right and to underwrite policies the right way, the better data that they have, the better that they can make underwriting decisions and make paybacks effectively into the right amount and obviously cross sell the different products and different areas, you know, stuff that they have in different business units across the business. So, you know, it is funny.

[00:17:15] I'm also thinking of a couple of cases back to this point that I've seen recently at big companies like MetLife and Roche, where I have meetings in the next week with people at those companies who report to the CEO and run the data strategy. So we talked about CIO driven data projects becoming CEO driven data products where they can actually demand high quality data for their business initiatives.

[00:17:40] There are now people like above the chief data officer, the chief data and analytics officer sitting on the executive teams responsible for that data strategy, which once again, just speaks to how important it is in the minds of executives at big, meaningful companies around the world. And as we said earlier in our conversation, without data, there isn't AI. The other big phrase I keep hearing is AI everywhere.

[00:18:04] So if we look ahead as AI becomes embedded in every enterprise system and those hundreds, if not thousands of applications, any advice that you'd give to leaders who want to avoid being stuck behind that data trust bottleneck, anything you'd pass on or any valuable takeaways for people listening? That expression, without data, there isn't AI. It's just emblazoned and you're seared in your mind right now. Yeah. Yeah. Yeah.

[00:18:31] And it was funny over the weekend, you know, we had friends over dinner, we're debating like the future of AI and what does it mean? And obviously I'm extremely bullish for the positive difference it can make in lots of different businesses and stuff like that. And you alluded to a little bit earlier, Neil, to structured and unstructured data. I mean, 90% of data out there is unstructured.

[00:18:51] And thinking about that and how do you apply metadata to unstructured data and how do you start driving quality against that unstructured data is another massive area that we're moving towards. So, yeah, you know, I always come back. So, I mean, I think about prioritization, pragmatism and, you know, progress, you know, there used to be the four P's of marketing and much of those are going to stick as the three P's of data management.

[00:19:16] But that prioritization is like find out what matters to the business the most and start there. You know, one of our strategic advisors, the chief data officer of Thermo Fisher talks about cargo ships and Ferraris. And a lot of data projects can look like cargo ships and they can take a long time.

[00:19:35] But if you can think about that project more like a Ferrari and think about the prioritization of that and connecting to a business initiative where there's real business pain, you're just going to get more engagement. It's going to be helpful. One thing about being pragmatic, that's that people process technology piece. And then the progress is metrics. Metrics, you know, we had customer advisory boards in London, Boston and in Sydney in the last two weeks.

[00:20:03] And one of the questions I ask folks is, hey, how does the chief data officer, how does the business know you're making progress? You've got all these programs, you've got data quality, data observability, data management, whatever it might be. But like three months from now, you're thinking red, yellow, green or red, amber, green, as they like to say over in the UK. Like, how do you know whether it's red, yellow or green? So really pushing yourself to be like, what does progress look like? And it's, you know, it's very difficult to measure everything spot on, but directionally, I think you can.

[00:20:33] So once again, prioritizing with the business, being very pragmatic from a people process technology perspective, and then really measuring that progress with metrics. I think it's all super important. Yeah, 100% with you. And don't worry, I'm not going to use that catchphrase again, my new catchphrase. But some of the other things that I learned last week is how businesses, governments and entire nations are starting to think bigger when it comes to AI.

[00:20:58] I was talking to the United Nations World Food Programme, how they're using AI to predict food shortages, climate disasters, recovery, etc. Serbia, here in Europe, they had transformed their exports 10x, again, by using AI. So there are so many opportunities here, aren't they, to get that data right and think bigger and solve real problems. Yeah, no, once again, over this sort of dinner conversation, I was reflecting on it Sunday.

[00:21:28] And, you know, if you think about accessing data, and you think about analyzing data, and you think about synthesizing data, AI can do those things in a way that's, you know, unprecedented. You can add, act on that data when you get into AI agents, and obviously that can get a little bit scary.

[00:21:51] But once again, the ability to access information, analyze it, structure it, act on it. I mean, my shoulder is hurting right now. You know, I went back into the boards. I'm going to go see Dr. Dan Quinn. That's one data point.

[00:22:11] But I am sure from the MRI that I'm going to get on Wednesday, if you pulled all that information from all the different MRIs that come from shoulders, no offense to Dr. Dan Quinn, because he's a friend of mine. But, like, I'm going to get a more precise answer around, like, why I can't lift my arm above my head, and why I can't take slap shots for a little while, and how long I'm going to be out, and stuff like that. So that's obviously in a medical area.

[00:22:37] But speaking about some of the examples that you gave before at T-Mobile, at Prudential, at MetLife, at Roche, I mean, gosh, when I was at Roche's corporate headquarters in Basel about a month ago, you know, when you think about patients, and you think about pharmaceuticals, and you think about curing diseases, and making people, allow people to live better and live longer, that's all a function of data, too.

[00:23:03] So, once again, I go back to what an exciting time it is, and, you know, exciting times come with lots of challenges, but, you know, our mission is to power a better future with data, and what I mean by that is really help our customers power a better future with data, and that opportunity is there for ourselves and almost every other company in the world. And that's a bold statement, but I really do believe it. And I think that is a powerful moment to end on today.

[00:23:28] But before I do let you go, we started the podcast today talking about your journey from professional hockey player in the NHL to leading a global software company. But, of course, in any vocation, none of us are able to achieve any degree of success without a little help along the way. Very often, if someone sees something, and us invests a little time, and us, and those people are probably blissfully unaware of the impact they've had on ourselves, our lives, and everything in between.

[00:23:55] So, is there a particular person or anyone that you'd like to give a little shout out to today that you're grateful towards that played a big part in your journey? Who would that be and why? Yeah, I know there's a bunch of former CEOs who I've worked with. And without a doubt, the biggest thing I learned when I became a CEO is how much I have to learn.

[00:24:18] And I remember I was dealing with Brad Spencer and Sean Dyke that are part of Door 2 executive coaching. And it was early in my career. I just showed up at Observe It where we spoke before. They're like, what's the strategy? What are you going to do? And I actually didn't know yet because I didn't know what the company was doing. I hadn't spoken to a bunch of their customers. I was still learning the space. And I remember them telling me, just tell the company that. It's okay to say, I don't know.

[00:24:48] And to have that level of intellectual honesty within an organization allows other people to say they don't know. So that was incredibly helpful. You know, the CEO of Rapid7, Corey Thomas, here in Boston, I worked for before. He's just an incredible leader. And, you know, I'm a big believer in any profession. Trying to speak to people that have been at it for about 10 years longer than you have. So they can still relate to your challenges. It's not 20 or 30 years.

[00:25:17] So they can relate to your challenges, relate to where you are and share those pieces of advice. It's super helpful. And then finally, I've been part of a CEO forum. Actually, this is a, I'll give this a little tidbit. Aaron Ayn, the former CEO of UKG, was a mentor of mine because his daughter worked at Rapid7. And I met with him shortly after becoming a CEO. And he's like, Mike, are you part of a CEO forum? I'm like, no, no, Aaron, I don't have time. I got to talk to the customer. I got to talk to the people. I don't have time.

[00:25:47] He's like, Mike, I've been at this for 25 years. I'm a member of two CEO forums. Join a CEO forum. And I was like, yes, sir. And from there, I went and I joined a CEO forum through Buzha Kukman. And it is something where we meet once a quarter for two days. It's eight other tech CEOs. And it's incredibly humbling, inspiring, educational, therapeutic.

[00:26:16] And just to be able to share the experiences with other people that are in the same seat or have been in that seat before you is incredible. And it's something that I try to share a lot. You know, I've shared a little bit through CEO Playbook, which John Belazare runs. A friend of mine in one of those CEO forums, another CEO. So lots of other CEOs. And as I said, I've made lots of mistakes. I'm going to make lots of mistakes. And I try to share those mistakes with others because that's what other people have done with me. And it's been so helpful. Wow.

[00:26:46] What a fantastic answer. And all those people you mentioned, a quick shout out to them. And they're probably blissfully unaware on the scale of impact that they've had on you and your career. But everything else we discussed today, maybe we want to find out a little bit more information there. Dig a little bit deeper. Contact you or your team. Where would you like to point everyone listening? Yeah, I mean, Atacama.com. Atacama rhymes with llama. Llamas are in the Chilean desert, which is the cleanest air in the world, so to speak,

[00:27:16] which is why the name Atacama came from. So lots of information on Atacama.com and obviously reach out to myself or anybody else in the leadership team that you see there or anybody else on the site. As I said, it's an incredibly exciting time to be trying to make sure people can trust their data and help manage that data in a comprehensive, efficient way.

[00:27:41] Because one thing we didn't talk about is data is exploding and sort of doubling every three years. The people managing that data aren't. It's the opposite. Meaning organizations want to have fewer people and they're pointing at AI and they're like, let's look at revenue per head. We can have fewer people doing all this work, yet data is exploding. So in that world, it's essential to have great data management tools that allow the same number

[00:28:09] of people to manage three times the data with three times the attention on that data. And probably three levels higher up paying attention than they'd ever seen before. So exciting to do that and super thankful to the folks at Atacama for all their hard work and our customers for sharing their feedback and helping us build great data management solutions and have great data quality products such that they can trust their data. And so much has changed in the last seven years since we spoke.

[00:28:39] But I think, as I said at the very beginning, you guys were doing this before. It was cool. And a global tech company helping enterprises clean, govern and trust data to make data AI actually work in the real world is just incredibly cool. And more important than ever, may I add, as well. But just thank you for the great work you're doing, shining a light on this and also sharing that great backstory as well for the NHL. We just can't leave it another seven years before we speak again,

[00:29:08] because I don't know how many shoulders you'll have left by that time. But we will get you back on next year, maybe. See how things are going. I have two fake hips, both of which are new since we spoke before. The sore shoulder. Extrapolating that out seven years does make me a little fearful. So, yes, let's definitely connect before seven years. If you don't mind, Neil, I would love to check back in and share or hear about some more great

[00:29:34] phrases that you've heard around the world and companies that are running their business and leveraging AI in a better way. So, great to connect. So, what stood out to you in this conversation? For me, I think it was Mike's point that the future of AI won't be decided by those with the biggest models. But, by those with the cleanest data and the courage to admit what they don't yet know. And his idea there that software is like hockey, it's a team sport that captures the spirit of how

[00:30:04] is a team sport, I think captures the spirit of how progress really happens. Yeah, it's through collaboration, discipline and willingness to improve every single day. So, what do you think? Are most organisations too focused on the shiny distraction of AI models and not enough on data trust? Love to hear your thoughts. Share them with me on socials at neilchughes or email techblogwriteroutlook.com

[00:30:31] or send me an audio message at techtalksnetwork.com. Let's keep this conversation going. But, a big thank you to Mike for joining me for the first time in seven years. Even bigger thank you to each and every one of you for tuning in. And let's do it all again tomorrow. Speak with you then. Bye for now.