Think AI Podcast - Why AI Fails Without Clean Data | Ep. 20 with Doug Saunders (Sara Lee)

Episode Date: September 8, 2026

🎙️ Why AI Fails Without Clean DataDoug Saunders has spent 28 years running technology in the businesses nobody posts about: waste, ice, coal, cold chain logistics, and now frozen cheesecake. He is CTO at Sara Lee Frozen Bakery, and his position on AI is refreshingly unpopular. You do not throw AI at a business for the sake of it. You find the business problem, you fix the data underneath it, and only then does the technology earn its budget.In this conversation with Dave Goyal, Doug walks through what that actually looked like in production, including a distribution problem solved with LiDAR sensors and predictive routing that took twenty to thirty million dollars of cost out of an ice company, and a customer service bot that absorbed 500,000 interactions a year.In this episode:00:59 A master's in Russian history, then golf course design, then CTO04:09 Walking into food manufacturing and finding the silos07:03 Why data has to come before AI08:49 Building a $180M company off QuickBooks10:25 Your LLM has no idea what your data means14:44 Don't throw AI at a business for business sake16:20 The ice box problem that saved $30 million21:30 Automating 500,000 customer interactions a year23:56 Automating to get rid of people is the wrong approach32:51 Trust versus governance, and the Excel experts who resist36:20 The end of Excel jockeying39:21 Fitting master data governance to your cultureIf you are being asked to "do something with AI" before your data is ready, this one will give you the language to push back. Subscribe, and tell me in the comments what your organization automated too early.🔗 Links & ResourcesDoug Saunders on LinkedIn: https://www.linkedin.com/in/dgsaundersDoug Saunders on Youtube: https://www.youtube.com/@SaundersdgSara Lee Frozen Bakery: https://saraleefrozenbakery.com/#DataStrategy #AIinManufacturing #SupplyChain #CTO #PrivateEquity

Is there any context to that data, are there any data dimensions? How is that data organized? And I can put an LOM on top of data, but it doesn't know what the data is. Welcome to the Think AI podcast. Each week, we talk about the most exciting AI research, tools, case studies, and more. I'm your host, Dave Goyer, and I've been working behind the scene in data and AI for over 30 years, whether you are an AI expert, skeptic, or something in between. This podcast is for you.

Welcome back to the Think Ki podcast. I'm Dave Goya. And my guest today is Doug Saunders, CTO at Satterley Frozen Bakery, the company behind one of the most recognizable names in American food. He has spent 28 years running information technology and some of the hardest, most physical businesses out there, waste and environment services,

eyes and coal chain logistics, to name a few. So Sarah Lee's founder built America's first fully automated bakery, and the company just came through a supply chain transformation. I would love to talk about that and anything that you would like to talk about. Welcome to the show. Thanks, Dave. I appreciate you having me, and yeah, we can take it any direction you'd like to go. Awesome.

And before I go further, I'm always intrigued. I always talk about the personal relationships and personal journey. I'm a disabled founder, spent 30 years in data and AI. What got you started in technology and raising up to a CTO level? What made you more exciting towards it? I have a non-traditional journey to technology. So I have a master's in Russian history, which obviously most people would think you're going to be in technology. And after the University of Alabama, where I got those degrees, I worked for the United States government for a

little bit, decided that line for living was not probably going to be long in my future. So moved on to another non-traditional technology, and I was golf course design and architecture. So I grew up around that and just did that because I didn't have much to do. And just by chance, we met the CIO at a startup company called Republic Industries, which at the time was Wayne Hizinga's company that after he sold Blockbuster that was starting what became Auto Nation. He was about to buy National and Alamo Renicar. He owned the sports teams in South Florida and he had started waste back in the 80s, but at this time he was starting up for Poet Services.

So it just by chance somebody took a chance on me because I could communicate and thought I could learn. I kind of use those principles now when I hire. And my journey started in the late 90s with a startup company around Republic Services and it's garbage and it's grown to what it's become today. That's funny and interesting. I live in Orange County and Republic Service is what we use today. And my wife and I are both disabled and they have some amazing services out there where they can take it from our, not from the car, but come to the house and take it. And I'm pretty grateful to them about that.

It's a good human-oriented focus out there. So loved. Yeah, that started with a gentleman name of Vince Taramina started that Orange County unit. So we bought the Taramina companies back in the late 90s, mid-90s, I guess, late 90s. And they've kind of continued that. They offered this top-notch, very personal service. It's very important to them. And obviously, Republicans owned them for 30-plus years and continue to do that good work.

Amazing. So let's jump on to field to factory. Most of your career was field services, RIS field services, waste, eyes, and cold chain. And correct me wherever I'm wrong. I'm dyslexic. So a lot of times I have mixed things up. So I want to make sure that I speak accurately.

You just stepped into food manufacturing. What surprised you most when you walked into the door? And what was new versus carried over? Because everyone who gets into the new company, that's always the thing you have to clean up. or something, but then you still have to prove yourself to do bigger and better while you are cleaning up. What was that look like for you? Yeah, so I've been with Sierra Lees since March. I've been in the private equity space for about

six years now coming out of Republican big publicly traded companies. And I think the lessons that I have learned in the bigger companies have translated well to the sense that, you know, These are all business problems you're solving. These aren't technology problems per se. And I think the biggest surprise, the complexity of all of this manufacturing, all the ingredients, all the supply chain. So there's a lot of, you know, we'll just start with what are we facing here with Sarah Lee. So the private equity sponsor, Colberg, has held this asset for about 70 years, pretty flat in terms of growth. And so brought in a new management team, including me, and really is to turn this around.

So, you know, working closely with our SLT, what are those things that really are going to help us create a better company? And I don't just mean by financial results, but really improve the culture. That's really important to our senior leadership team. I think it starts there with our people, making sure people, you know, we kind of go by the motto of, you know, trust, transparency, collaboration. We do that really from top down. So it was really coming in and setting that agenda. Second, really, where were the problems? What was happening?

And I think you kind of saw some of this in the supply chain and the manufacturing side. Yes, the tech stack is a little challenging in that it's a very siloed company. And say, for example, sales says, I need data, my own data. I'm going to go out and buy applications and put applications together. and I'm going to say whether it's around price or, you know, we're trying to do some analytics. And, you know, most of our businesses is through food service, U.S. Food, Cisco. They kind of created their own data environment, not just sales, ops did it, finance did it, you know, resource and development did it, probably because they didn't really have a strong enough

tech leader to help put some sort of enterprise architecture together. So while it's not really surprising, it's overcoming those sites. and building trusts within the organization that I'm going to put an enterprise data platform together. We're going to add capabilities like predictive analytics and AI to that, that data platform to really help the business make quick decisions. And probably like most companies, at least we see this in private equity, you know, how well you use your data is really going to depend on how fast you can exit at its at a success. And so, for example, for us, how do we not use data to look what happened months ago,

but how do we use data to figure out what's going to happen next? And so that's been some of the challenges that we've been trying to address sincerely. That's amazing. And I constantly talk about, I have a running blog and this podcast. I also write on Forbes Council. And one of the thing that we are constantly talking about, which is you are also leaning towards data before AI. A lot of people get really excited about AI,

But for enterprise, for corporations, companies, even mid-sized, the data is far more important, right? Because if you don't have the right data, AI will only accelerate the bad decisions or bad insights out of it. It's not going to help you anymore. So that's one thing that always intrigues me. And, you know, we sell data and AI to mid-sized manufacturers being a Microsoft partner. So I'm looking from outsider perspective. But your insights are pretty valuable what you're explaining here, so I appreciate that. From that point, let's talk about the data problem.

I forgot which year, but the CIO magazine was saying every other project, which is like 50% of the project are failing due to data quality problem. Now, when we started thinking deeply, most manufacturers should have an ERP. So they should not generally have data problem, right? because it should be validated when there is bad data, it could be due to a few things that you are missing certain data elements, you're missing some historical thing. Or example, you have a feel for, you know, state and which has like CA or California,

some of those things, right? You'll call that as a data quality, but there are other examples also. You have worked 28 years across this field services and manufacturing. What is the one worst data quality problem you inherited? and what it took to actually fix it, if you are willing to shape. Yeah, well, so it's funny that I've been in all kind of businesses, and so, you know, when I think you work with it,

I think you got two problems, probably the most unique challenge that I was very early on with a environmental service company in 2020, that was a startup. And there was no system. I was employee number one. It was about a $180 million company. that was basically run on QuickBooks.

So the unique challenge there is, and I've never had to face this, is I've got to find a system and build something from the ground off. And do it, you know, in a private equity setting that's really demanding, you know, especially back then. This is low interest rates, pre-inflation, deals returning fast, the turn rates for three to five years. How fast can you put in an enterprise system, whether that's a big ERP or best of breed or whatever it was?

was to actually grow it. So I think that was the most unique challenge. And there we were able to put in Netsuite as kind of the financial backing. It was a field service company. So we went more to lead the cash on the Salesforce side. So not just CRM, but service cloud, CPQ. We did a lot of government contracts. And we did field service as well. So that whole life cycle from really lead to the invoicing was all done in Salesforce that then fed the financials in New York. that gave us that data foundation then that we could turn into insights. So I'm used to walking in to, you know, like Sarah Lee, we do use NetSuite, for example. So I'm used to walking in, okay, I don't necessarily have to build out a system.

But I do have to worry about some unique things in the days of AI and predictive. And that is great. I have data. And most of our data, 80% of our data really kind of comes from NetSuite. But is there any context to that data? Are there any data dimensions? how is that data organized? And that's really, I can put an LLM on top of data, but it doesn't know what the data is. And so we've had to really start a project where we're going through and building out that context around that data.

So we can't ask its specific questions and give us four or five most profitable customers. Where are we losing margin? Where in the distribution area are we? And by giving it that context. So we're kind of, you know, just have started that we have a kind of, you know, a North Star data project that we want to build. And we're kind of in the beginning of that. I mean, obviously, only been here about four or five months and, you know, starting that journey. But it really started with orientizing that data and giving it the proper business context.

So now as we build out, okay, yeah, we can build our KPIs, but now we can get into building these co-pilots, you know, for lack of a better word, to allow the finance team to look at financial data and ask it specific questions and actually get an answer. And we can also start predicting the future. We can use the data that we have and create models that actually help us predict demand, say from some of our customers. And now we can add in insights that, you know, traditionally we haven't used before, things like could be anything from restaurants, economics, right? Meaning what's really happening in the restaurant industry? What's happening in the We sell a lot to food service.

What's happening in the hospital? The schools, all these places food service sales, we can add all these insights in. We can use industry data. We can use partner data. So, for example, we can actually get some data from Cisco. Now we can create, we can start helping model data, you know, for demand and supply chain to say in these particular, say, economic conditions, weather conditions, restaurant, you know, consumer trends.

how is this going to or how is this going to affect our revenue? We can then plan that way. And so that's kind of the model that we're looking to. And it really, to your point, starts with the data, the right context around it, the right dimensions around it. So these AI and predictive tools can use it well. That's pretty interesting and amazing.

And one thing I love what you've been talking about, most CTOs get hung up with due respect to them on technology. part and you're not saying a lot on those technologies really thinking from that business challenge and business problem. In fact, we're working with one of the hip, you know, a burger chain here in Southern California and similar thing. And what happened with them is, so we got engaged in this data project. I'm really good in data architecture and even though my team most X Microsoft, but I love to get into initial one where I can work with CIO's CEOs, CEOs, CEOs. and CFOs. So we worked with them. And then we realized that their point of sale system is really bad.

And we got up to a level that they halted the project. We happily did that. And now they are changing their point of safe system because it's not capturing the right information. Like you're saying, you know, the whole unit economics, you know, which stores are working, which distributor mechanism is working and whatnot. We've been using Microsoft fabric to name the technology. with the IQ, but it's not important, right? Technology is not important. The point there is what was the real challenge? And their challenge was even bigger,

not even capturing the right data. How do I really produce anything when the data is not there? So this project really surfaced that really fast. They wanted to jump onto the AI right away. So we first went to data, data quality, and up to a level that a data source is bad. And the example you mentioned, kind of not bad, but at least not making it

bad, you know, going into the right direction, to the right foundation towards the AI, isn't it? Yeah, I mean, you, I've seen peers that have really got hung up on our company needs AI. You know, and fine, if you want to, if you're talking about like personal productivity, like if you want to throw out some chaty BT or clot or whatever your flavor is, right, great. I mean, that can help people be a lot more productive. I don't know that I want people making, to our point, about, you know, using that for enterprise-based decisions simply off a chat GPT model, you know, without some testing, et cetera, because, again, it's about the data quality behind it.

But I don't know that you just throw AI at a business for business sake. It's, it really does have to have an ROI. And it has to, you know, I've been in companies where I, we just left a company as a package of ice company called Arctic Glacier. You probably have that in Southern California as well. And, you know, that industry, you know, was really interesting in the sense that it wasn't about just going out and finding AI, but really what business problem is trying to solve. And in this case, we had a big distribution challenge where we were running routes. And if you think about what's outside of 7-Eleven, you open up that ice thing and you pull out your ice and you go in and pay or vice versa, when we make deliveries, we don't know how full that is, right?

And so we use, and so we would be running routes and you'd open it up, you'd only put one bag in, the next stop one bag, the next stop one bag, then you put 300 bags in. And that's where you want to focus on that. So we use some IOT technology using some LiDAR that actually gave us the inventory levels of those ice boxes, which then turned into a, we went to a predictive data analytics model that said, here is when we should actually go deliver. And then it would auto route that. So it used them a general, AI, eugenic AI, you know, Python scripts, some things like that, that would really then turn that into an order and a delivery. And so we really were able to save $20, $30 million, you know, and just labor, truck costs, logistics costs that really drove that cost of utilization. So to me, it's about finding use cases that technology can help solve for the business versus saying, I'm going to take a particular piece of technology, AI, whatever it is, and just start throwing it at the organization. How do you measure those results? How do you get results out of the organization? Because in private equity, as you will know, they care about EBITA. You know, you learn very quickly what, you know, if you're coming from a different world, how important that profitability is. It's not about volume. It's not about much revenue you have as much as much as it is, is how profitable is that revenue. So that's where I think technology can really help, especially in my world of private equity now.

Well, that's absolutely true. I had two exits. I had built nine businesses. Five of them were miserable failure. But then I had two great exits. And this company, KIA, is three times INC, $5,000 awardee, which is totally focusing on the BITA. And, you know, I studied CFA. So I understand that value.

And you're working with the P.E-owned companies. That's the key thing they are looking at. Everything else is, it's important, but it's not as important as that one. And generally, it's a good matrix from leading and lagging indicator, right? So, you know, if it's a lagging indicator, meaning it's showing the performance, but even from leading, if you are looking at month to money, a bit, that will give you a good indicator of what's going on of the company. Then you can drill down on revenue or expenses or anything like that.

And I wanted to ask you, you know, you mentioned an AI success that, led to an exit. But I guess that's the example you just mentioned if you want to expand on it. Yeah, I think that it's, you know, we, and it's funny is that it's, it's not so much that I'm, say, bringing a solution is you have to, it starts, I mentioned the culture at the kind of the beginning of the conversation. And where we've had the most success deploying technology initiatives that really help, you know, growth in EBITA and all those fun things that private equity likes is, is because the culture is right for that. I've been lucky to work with a lot of great CEOs. One of the best I've ever worked for. I work for now and I work for in our last company, Pete Laporte, who really

allows the senior leadership team to build this amazing culture. We really work at it, right? We go do off-sites with an executive coach. I mean, I've learned so much about myself. I was telling my wife, when I left Arnda Glacier, I left, you know, with a lot of great technical. technology success, but I left being a better person because of this work. And so that culture has led to a lot of success because it's not just about the technology. It's about how well you can work across teams. And you really have to learn the business. You have to be a technologist for sure, but you really do have to understand how the business works. And so I had a lot of great partners at Arta Glacier that when we talked about this complex distribution problem, how are we going to

solved it. And there was a lot of different ideas that that team came up with and we worked really hard together. We kind of all bought into the same mission that it's okay to kind of try, experiment, and fail. Obviously, we did things at scale, meaning we piloted things first, we tested things first. We found great partners that were outside our company. Our IT team at Arctic Glacier, I literally went from 30 to 9, right? So I shrunk the team and got a lot more done because we had great outside partners. Now, that's not for every company, but in that situation, it worked well. And so really the solution that we had come up with around, you know, really reducing our ability

to, you know, cut out trucks, cut out labor on the distribution side, auto route, auto order is amazing. Some of the other things we did was on the customer service side that our customers, you know, yes, we, Walmarts and Costco's, we send a lot. of ice to, but we also had a lot of like small businesses that were, you know, one or two gas stations or, you know, we're selling ice somewhere to campground. And they'd have to call in and order ice and let us, hey, I've got four bags left. It's my busy season. I'm going to have no ice soon. It was very problematic. So we took about 500,000 interactions a year where it was just I want to,

I want to order ice or I want to know where my order was. And so again, using someogenic AI, we actually automated all that. So we came up with a bot on the website and our mobile app called Glen Glacier that we actually would allow customers interacting that could get all this data at their fingertips. So again, we went from 54 agents, I think, down to 9 or 10 or something like that. And they were focused on the bigger problems or routing things to sales. They didn't have to deal with these interactions. So I think those are more examples of, of where technology, understanding the business problem, partnering with, you know, our customer service team.

So Allison Dennett, who ran customer service there, was probably a little skeptical at first, okay, this isn't going to work, but she partnered with us, right? She was there testing on the front lines. She made suggestions. She was a great partner. Same with the logistics team.

So, again, it's that business partnership. It's the right culture. It's the business partnership. And then it's applying the technology. and you can get some great successes that way. This is an amazing example. And, you know, Gartner, since beginning or several years, 26 years I've seen, they talk about these four quadrants or four levels, right?

Descriptive, diagnostic, predictive, prescriptive. Your example is fitting right there where you actually achieved a prescriptive model by being a little more innovative, having an IOT in place, figuring it out what is needed. and where it is needed without raising the cost, but still giving a lot of value out of it. So amazing success story.

One thing you touched upon, and that will lead me to my next question. So I also wrote a book on real-time business intelligence, and few things I put it in sequence, people data, people data culture, technology, and automation. And I personally see that in that order. And automation, I put it at the last, because if these things are not in place,

automation doesn't help is my belief. I want to pick your brain. So one of the phrase I have is automate the decision and keep the judgment. If the decision is that the chain would be automated and the human is still needed, what is your thinking behind it, where the handoff you see to automation

or to a tool or not see? Yeah, that's an interesting question. I think automation certainly has its place. There's a lot of things that I think especially in, you know, Sarah Lee, for example, that our teams, it's not about reducing headcount. You know, everybody hears automation, you're going to reduce headcount. I mean, sometimes that, that works. And, you know, in a lot of cases, it's not. And so, like, it's Sarah Lee, for example, I think 82% of our orders come into EDI, for example. And we have trading partners, you know,

small, not the Cisco's of the world, but smaller food service companies that may or may not have EDI, right? And so all of that is manually done. Now, our customer service team, you know, should be dealing with issues and problems, not I've got, I want, you know, 500 cheesecakes per se, right? That's, that's an easy automation. And so in those cases, what we've done is, you know, started to automate some of that, that order process that allows our customer service agents to focus on really what they're supposed to be focusing on is handling problems, helping balance, looking at inventory levels and working with our customers on order management and not so much just this laborious work. So I think sometimes by taking out the low-hanging

fruit that people generally don't like doing, you're allowing those individuals to really focus on where you need the human decision and the human interaction with a customer, and it's so critically important. I think that's really where I love to do automation, because like I said, like even in an Arctic where we, yeah, we went from 50 to 10, we didn't need when you get rid of so many interactions, you don't need all these people. And that's kind of the unfortunate side of the business and some automation. You do hear that. If your sole goal, to automate is to get rid of people. I think that's kind of the wrong approach. I think what you're trying to do is automate where you can, where it makes sense for the biggest value, and that value may not necessarily be just headcount. It could possibly be in this case.

Our customers get better attention. They're better account reps. They're better at managing the group. They understand what that customer needs and wants, which help in turn helps our demand planning. Right. So now we know what they want. So now we can turn that. into a supply chain order and then a production schedule, that's more fitted versus a little bit more of a guess. So I do think there is some value very much so in automating and still allowing the human to have some decision. And not just decision, but just human interaction, especially with a customer. You said it well.

And that's why we were putting people in automation at the last because, you know, if you do it reverse, do the automation first to get rid of people. It's not going to work because people have skills. The thing that is more repeatable is where you want to do automation so that you can free them up also and get better value to the customer. You like you said. This also leads me to so this is more on the technology part.

Going back to the business again, a supply chain transformation, you came to the far side of it. What did you find overall? I worked with a lot of bit-sized companies like our day gear companies and fall gear companies and things like that. And everybody had a different challenge. So why supply chain looks pretty, Wadilla, it's not, it's really industry, and segment dependent.

What's your findings? And then from supply chain, what's your next chapter? What do you want to tackle the most or love to tackle the most? Yeah, I think what's really interesting about our current situation that was a little bit surprising to me,

in the supply chain side is we have a lot of the great tools. And a lot of them are either overlapping. So we have really three demand and supply plane planning tools, whether it's the Netsuite capability, it's the Atlas, the John Dalt product, whether it's Optiva. There's all this overlap. We have CRM tools, right, that are tell us, which is food service industry kind of specific, right?

it ties into blacksmith, which does trade deductions, but NetSuite has a CRM. So we have a lot of this, you know, capability in the ERP that we do not leverage. And so I think the trick is not so much, you know, we've got really smart supply chain people. Matt Savage, our chief supply chain officer, brilliant, like he's like a human computer.

But we have all these systems that aren't really optimized. And so I don't know that it's necessary. there are some supply chain challenges, don't get me wrong. I mean, like, are we getting the right price? I mean, just earlier this year, there was this aluminum storage. It might have been geopolitical because there's so much aluminum product that's coming out of maybe the Middle East, right, that the Iran consulate kind of, you know, and so how do you, you can't make pie plates, right? So what do you do? When you make a bunch of pies and you can't make pie plates, that's a challenge, right? But really for me, I think it's, there's really two focuses,

is how do you optimize the technology investment? You really can drive some costs down. I don't want to lose capability, but I also just don't need to pay an extra 300 grand a year just because there's one little functionality that doesn't. So really rationalizing that, coming up with the best architecture, because with all these disparate systems,

now you've got a data integration problem, then once again, a data quality problem, right? So it all kind of reaps back to the data. And then I think the second one is, is really the demand and supply planning is, you know, from my perspective, we're working to get closely with those teams. And it's part of the optimization as well. But if you're taking a demand that comes from historical sales and then says, okay, we're going to take some input from, say, the sales teams and come up with a demand plan, is that really in today's world the best way to look at how you do demand type planning?

There's so much data sources out there that could help us predict this better that then minimizes our investment in the supply chain. We can tie that in with inventory levels and then with our distribution partners, all the 3PLs that we're using, right, and come up with a better plan. Again, that gets back to data, right, first, right, and getting the data so you can analyze it and get the data in one place and one enterprise data intelligence platform. And so that's really our goal. So I think the two biggest surprises are really coming in and like, why do we own so many applications? And then two, how do we really streamline and make the forecasting part of supply chain more accurate? Well, that's amazing. I am a big follower of both Bill Edmund and Ralph Kimball.

And these two gentlemen have taught us a whole lot, you know, get to work with their organizations at Disney and other places. and you know the kind of thing you could do to put data in one place, whether you call it data warehouse, data repository, lakehouse, data lake house, whatever word you want to use, putting data at a homogeneous place where you can get everything connected and get the answers is amazing. A lot of hard work which a lot of people don't or not able to see to get to the level of prediction of prescription on the data and analytics part.

So I commend you on that. I want to switch gear on a little bit of organization challenges. So there are two parts in data world and AI even. One is earning a trust because people working on Excel sheet, they become the champion. And when you try to convert into, let's say, a report in power, BI or something else, they'll always, always challenge. Rightly so because a lot of that logic sometimes live in their head. Sometimes that logic is being, I don't want to call it twisted, but it is.

because, you know, based on how they want to report it, and not in a bad way, but then they are doing that calculation, but it's not really documented. There is no data dictionary or anything like that. So when you build a data system and data architecture, you are bringing all that in one place, right? So gaining trust is one thing, but then they are losing control,

and you want to control as the tech leader in the IT sector. So now you want to build governance. So trust versus governance in my mind is too competing, thing. How have you dealt with it and how did you resolve that? Yeah. So in larger companies is a much different challenge than at least it used to be. It's been a while obviously since I've worked in, you know, 15, 20 billion dollar companies. There it's a lot slower process, right? It's, it is building. It's, it goes back to my political days, right? When you're, you know, you're trying to work across the aisle in a political

scenario and build consensus to get something done. There's a lot more of that. When you get into private equity, what I've found is a lot more nimble and quick. I think what helps in private equity is that you have this goal, especially where Pete and his team have come in in the last two companies at both Sierra Leen Arctic, was there's a mission. Really the goal is, you know, the private equity sponsor has sold this, you know, has bought this company seven, six, seven, eight years ago, and they're ready to sell. It's a long time to be in a fund for a private equity company, and it's a little bit of a turnaround. So there's this little bit of an urgency to say, we can't keep doing things the same way. And as you get embedded that and you start,

you know, turning that into part of the culture, which takes time, it, it starts. It, it's to become a little easier to affect this change that you just described. So say, for example, at Arctic or Sarah Lee, you know, as you get to that six months to a year mark, these walls and these silos and these Excel experts start to realize that these things have to change. And there's a little bit of a different pressure than, say, at Republic, where you'd have to go to 78 meetings and build consensus and we want to make this change. and very different. And so, yes, I think you still need to build trust with the business teams.

You have to partner. You have to look for automation. Back to that, our infamous word there. But, you know, we have a lot of in our finance team, there's a lot of, instead of using NetSuite, which is a, you know, pretty robust ERP, instead of using it to its fullest and using its capabilities, we tend to do a lot more in Excel. The sales team, when we do pricing, for example, when we look at, okay, what are our pricing initiatives, you know, around certain skews or customers, it's all done in Excel. And I think what you have to prove out is especially using data in a kind of an automated way and now with the strength of some of the LLMs.

So, for example, in this enterprise data initiative we're working on, we're starting to model those two examples as kind of pilots where we can look at pricing and profitability automated through an LLM on top of this data. So we're using in this case, data bricks that sits to Azure that allows us to then, you know, query using an LLM, getting real time almost information. So we first got to Sarah Lee, you'd sit in a meeting and we'd have a pricing question about a customer. I'll be back in about a week after I go do my Excel, right, my Excel jockeying. Well, you know, very soon, we're just going to be able to sit in a meeting and ask the question of the LLM and it'll spit all this out, you know, with a high degree of accuracy. Same thing in finance, right? So we're going to be able to automate a lot

of things, not only in the data enterprise platform, but this is an optimization of NetSuite that just says, look, we may need to change some of our processes around a little bit, but we're going to be able to automate things like, you know, cash flow forecasts instead of it being in Excel, all these things that you can do systematically. And I think our CFO brought up a good point is that, hey, look, teams need to know how to build these manual first. They need to understand the process first. And as we layer in technology to it, we can then see if that technology supports that process, we need to modify the process, et cetera, because we don't want to go heavy customization if we can avoid it.

So those are really kind of the things that I kind of do. I mean, now, to do all that, you've got to have a lot of trust within the organization and a lot of support, especially from the top down. And so we've tried to create this mantra on what we call the SL senior leadership team, the SLT, is that the SLT is kind of one team. It's not Doug and his IT team, and I'm worried about that, and somebody else is

worried about sales, we're worried about what's best for the business. So if there truly is a consensus on that SLT that these are the areas that really do need some of this automation, lack of manual process, getting rid of the Excel portion, we tackle that together. So I think, yes, it's technology, but it is really about the relationships and the trust and the transparency of what you're trying to do that kind of make these things work. No, that's pretty amazing. And keeping that focus, keeping a balance, and it is also going back to the culture point you mentioned.

Every organization based on the size and their style of working, leadership, etc., etc., creates that culture, and the governance part then reduces because of the culture and, you know, taking ownership, taking responsibility, the SLC, SLT example you mentioned. That's pretty amazing setup you got there. One last question before we wrap up. You know, some people think this is an old school thinking.

Again, me spent 30 years like you. So I have some old school thinking still, even though I'm heavily active in AI. Master data management in my book is pretty high. Most of the data browsing world, people get by with dimensions, but that's not it because finding the authoritative source of data, which system it lives in, you already mentioned NetSuite and CRM and others. and then pulling what attributes from where and then creating the data steward,

the data champion so that you can keep the data hygiene really high. How did you deal with in these organizations? What kind of thinking you applied for master data management like PIM or SIM or just dimensions? Walk me through with that thinking behind it. Yeah, I think every culture is a little bit different, right? You know, in the ice business and maybe the cheesecake business, people don't think this isn't medical records or, you know, a bank. And so I think you can go,

it's kind of like running a project management office. I can go McKinsey level project management or I can go what I would call tree hugger, you know, you know, PMO, or I can just kind of go PMO light. And I think, you know, it's the same concept here. And I think what's kind of what's happening is for it to work and work well, it has to work within the culture of the organization. So Sarah Lee, for example, okay, so really what is this going to be? So what we've really, so I came in and what's interesting is going even to step further back, you know, without even talking about source data, there was really no, okay, here's the thing around data retention or classification, things that are kind of IT 101,

that not even existed. So, okay, we got to start somewhere. And where are some of the unique tools that could help us do that? So I'm big in, I love looking at startups. So, for example, not a little off topic here, but we're looking at a Stanford startup. These kids are right out of Stanford, you know, MBAs, and they had this trade deduction software using AI that could help us with which reductions, right? So I'm big on that. We found something that does data class AI, data classification that really can go out and look at all of our records and these things.

And so I think it starts here with the base. We put together a team of IT and business leaders. Every function has a data champion. And so we're starting this data governance kind of council. I've never really done that anywhere before. And I felt the culture here, though, is this collaborative way of doing it. And so they're putting that, we're putting that that's been put together. It's kind of in its infancy.

And it's looking at what data attributes are really the most important for each section of the business. So if it's manufacturing and it's OEE and its attainment and it's all these things in manufacturing or if it's in finance, we kind of have the experts in place now to really identify what are these attributes that are super important so that we can pull these in, not saying we're going to limit it to that. You know, nowadays you can actually obviously take, you know, all your data and lake it and, you know, and all these things.

And so it's great. But I think it really comes down to that. that. You know, what will, what methodology you typically will use is based on the culture of the company. At Arctic, nobody on the business side had any interest in it. And so we would use more of a, okay, IT-driven, very technical, I mean, think of things like Informatica and all these, you know, old-school type system, whether that's old-school or not, I don't know, but, you know, the type ways of doing it, we kind of had to do it that way. And so they can also get so, like at Republic when I was there, and this has gosh, been 11, 12 years, so who knows what they're doing now,

it was super high governance. It was like congressional committees, right, that were doing this. So I'm a big believer in fitting it to the culture, doing what's practical and make it very nimble, quick, and really can adjust on the fly. That's really good. And, you know, a funny comment about the old school technologies I came from Ebnacio, informatica, cognos, they were heavy proponent of, you know, these components. But what you're describing from that collaboration angle, you don't have to know, you know, how the data, a process, or report, or IT governance work, what IT you already know. But a lot of other things you don't need to know. As long as people collaborate together and in this era, you research through

different AI models, you can get to a pretty good solution on your own and bright minds like Stanford as an example. I also work with UCLA. They have really bright minds and if there are bright minds, they'll figure out something so amazing that you won't even think about it. So an applaud to you on to that. And with that, Doug, thank you. I really respect the way you have spent your career, which is the unglamorous work. I have done it myself so I can see the value. You know, data and AI actually have to work together. And if they don't, a wrong for in your case is, you know, waste on a loading dock, right? It's not just a bad slide. So it's either missed revenue, most opportunity, or more expenses. And you have done it well.

So I appreciate you on that. And thank you everyone for listening, Doug Saunders. I really thank you enough here to connect with me. Thank you. I really appreciate the time. Enjoyed the conversation. And yeah,

feel free for, you know, to anybody wants to reach out to me for anything. LinkedIn's certainly the best way. Thank you, Doug. Thank you. You have been listening to Think Yai podcast with Dave. Take one idea from this episode and turn it into action.