Kelly Kohlleffel (00:00)

Hi folks, welcome in to the Fivetran Data Podcast. I'm Kelly Coleffel, your host. On the show today, we bring you insights from top experts across the data community covering AI, machine learning, enterprise data, analytics, and more. And today we're thrilled to have Paul Bruffett join us. He currently serves as VP of Data and Analytics at Jack in the Box. Prior to that, Paul held leadership roles at Starbucks, Accenture, Sonic, and Devon Energy. Paul, welcome in. Great to have you.

Paul Bruffett (00:28)

Thanks. Excited to be here, Kelly.

Kelly Kohlleffel (00:30)

You and I, we caught up a little bit at Ignite a few months ago and had a lot of fun there and chatted a little bit about some of the changes that have happened in the data space since you and I connected up a few years ago. There's a lot that's gone on. Tell me what's been happening with you recently here and just kind of catch me up on Paul in general and data with Paul.

Paul Bruffett (00:56)

Yeah. So, uh, like you alluded to in your intro currently, I'm VP of data and analytics at Jack in the box. We're actually called Out of the Box Brands. We own Jack in the Box and Del Taco. Uh, so got your choice of the cuisine you'd like from us, but, uh, and I'm being here about. That's right. 

Kelly Kohlleffel (01:11)

Like it. Jack in the Box has great tacos, too, by the way.

Paul Bruffett (01:13)

Yes. Plug plug for Dell. Yeah, exactly. Yes. Um, and, uh, so before, you know, I've been here about six months in the role and definitely look forward to talking about the great work the team and I are doing. 

Before that, I was at Accenture for several years doing everything data and analytics as Global Retail Chief Data Scientist focused a lot on genAI and in some of that, especially, you know, it's a huge area of focus, both for Accenture and for the industry. And then as you noted before that, I was at Starbucks for several years, same kind of work, although less gen AI at the time. Uh, and before that, only guessed, but really data and analytics across a few different industries. Yeah.

Kelly Kohlleffel (01:53)

I love your background too, because you have that GSI background with Accenture. You see a lot of things that way. You've got the industrial space with Devon, and then you've got these retail, major, major retail brands as well that everybody knows and loves. When you look at your background, Paul, and the perspectives that you have on data, how have each of those industries and those experiences from both the services side, the retail side, the industrial side, of shaped how you feel about data, where you use data, and where data can really impact an organization that you're with.

Paul Bruffett (02:27)

Yeah. I mean, as you noted, worked across a number of different industries. I think the thing that's common across them is everybody's hungry for data. Right. I mean, I remember, it was a little bit fashionable at the time, but whenever I was in oil and gas, there was a trend of calling yourselves a company that's really a data company that does X, right? So a data company or technology company, the drills for oil or transportation company that's died down a little bit. But I think there was a lot of truth to that, and there remains a lot of truth to that. Right.

And in oil and gas, folks would start their day with the reports that we would generate about production and how wells are performing, how the business is doing. And it's really the same in quick service restaurants. It's the same in retail, right? My stakeholders at Jack in the Box, they start their day with the reports of how did we do today versus yesterday versus this period last year, you know? And so that's key to our understanding of the business as well as making all the decisions about staffing and inventory and everything, right?

And again, I think you see that consistently across industries. I think obviously there are a lot of differences, especially in kind of competitive pressures across some of these. QSR is really competitive right now. I think you've seen that in a number of different areas, which leads us to some different focus areas. Right now it's very much around how do we, how do we manage margin? How do we contain costs? How do we control inflation, both labor, you know, our materials and even the building, right. And, and redesigning restaurants. And so, obviously a lot of industry-specific nuance, but I think a lot of commonality across them where you see folks continuing to invest in data, even as we see some of these businesses paying a lot of attention to costs and capital allocation.

Kelly Kohlleffel (04:10)

Yeah, absolutely. And the reliance on data across industries that you talked about, it seems that now with what we see personally, and you think about data personally on your phone, laptop, Netflix, Spotify, whatever it is, I need that right now. And so as reliant as we are, my guess is you just get more and more demands to give me that data now, give me more of it. I need it from more areas across the business. And I want it in a way that's easy to consume. Don't make me work for it. How has that shifted over the years, this need for immediate data and more consumable data?

Paul Bruffett (04:54)

Yeah, I mean, I think there are a few planks to that. I think one of the key ones is what you identify where there's an ever-increasing appetite for data. And I think in economics, there's this term “induced demand” where it's the same reason you never have enough roads, right? We build more roads. You see these massive roads in Los Angeles or in Dallas, in Texas, you know, and it's like, well, they're still full, right? You add three lanes and it just fills up again, right? And that's the concept of induced demand where you can't build your way out of demand for services like roads. 

And I would argue data is the same, right? I've worked for a number of organizations. Starbucks grew the data team massively during my tenure there. And it didn't, it's never like, okay, we grew the team. We've answered every question we have about customer. We're good. We're good here, guys. Move on. You know, instead it was, well, we answered a question, and now there's an increased appetite from the business. They want to know something new or they want to do something new with that data to drive innovation in our services or our offerings or our products, right? Or the digital experience, right?

And so you increase the demand for data in my experience increases with the amount of data that you provision to your stakeholders. And so it's, it's kind of this induced demand, which is great. It's great for data teams. And I think it's great as a data leader because it shows the impact that business is having and usually results in increased investment. But it also means that you can't grow your way out from a head count or from a staffing or resourcing perspective from the data asks. Again, I've never been in a business that said we're really happy with our data. It's clean.

We've answered every question the business had, right? Regardless of the team, the team size. so kind of to your point about data access, right? I'm huge. I'm a big believer in data democratization, although it's very challenging, and I'm a big believer in giving folks the access to information to both consume it in the format they want, right? Do they want to consume it on mobile? Do they want to consume it on web in their email? A lot of my folks still love getting emailed, you know, copies of dashboards. I think that's great. If that's how they want the data, that's how they get the data, right?

But, but it's, it's kind of that combination, right? I've given people the tools to really interact with it and ask questions of it in the way, the format, kind of the right time, that suits their needs.

Kelly Kohlleffel (07:06)

Yeah. And you also talked about data democratization. Talk to me about what that means to you, what that means today to your team. How has that, because that, that terms been around a long time. And I think, and you mentioned this a bit. mean, that's, that's a, it's a tough thing to ultimately achieve. Do you feel like we're there today? Talk to me about data democratization for a couple of minutes.

Paul Bruffett (07:26)

I think there are a few planks to that, right? I mean, I'll be totally honest as I haven't, I've worked at companies I've worked at, we've, we've ended in different spots, right? One of the companies that it seemed like it worked the best at was actually Devon Energy. Maybe that was some nuance of the company, relatively small, given the size of the company, relatively small employee head count that is given the size of the company, very engineering focused, right? It's oil and gas upstream oil and gas, but we had folks we brought in, uh, Snowflake, we brought in Databricks, and data replication technologies to get the data landed, right? 

And then really built a mechanism for folks to interact with the data and build their own insights. And what was the most, I'd say gratifying and compelling about that is we had folks from one business unit, right? One base, and that would build insights that were actually embedded there. And then we scaled it to all of the other business units because ultimately they're the ones that are closest to it, right? They understand the needs of their business the best is really difficult to translate that to a data team or a data science team or a decision science team effectively and have it come out. And so that was great because we really gave them the data, and we gave it in real time, right? A lot like we're doing with Fivetran here at Jack in the Box. And that also unlocked new use cases, by the way, is to have it up to date in real time, not nightly, not, you know, way after the fact. And then we gave them a playground to really get the tooling that they needed, you know, not just, oh, a visualization or dashboard, but get into Snowflake and build your own.

You just get into Databricks, build your own models, right? And then we operationalized it, but it tends to be a challenge to actually come back and refactor and live with it. There's the governance challenge that we can always talk about. I found other organizations, it can be a little tougher. Maybe it's the nature of the organization. Maybe it's the nature of the kind of work. But to me, really the pieces and the pieces we're putting together at Jack in the Box, Del Taco, is to give folks kind of that continuum, right? Hey, if you're a really technical user, you can get data science notebooks. You can get a development environment, a coding environment and the data, right? Or a SQL interface like a snowflake, you know, and, or the dashboarding interface. But ultimately we bring all the data and, you know, are, being much truer to the extract and load and then transform, where in the past, when we were in the data warehouse environment, right? We had to do a lot of that curation upfront. And we left a lot on the cutting room floor that really precluded the kind of work that we're talking about here.

Kelly Kohlleffel (09:47)

Hmm. I love that. mean, that, really speaks to that data democratization that you mentioned in you. You talked about, Jack in the box. You have the Del Taco brand. You've been at Sonic. You've been at Starbucks. When explore for a minute, just food and beverage in general, that is a, it's a, it's a challenge. It's a fun industry, challenging industry. You talked about margin pressure, all these kinds of things.

Anything that comes to mind when you think about common obstacles, maybe across those companies, maybe it's even more so today for data integration, data analysis, building new data products, tools, platforms, approaches, ways of, know, kind of even structuring the organization you found particularly effective and working well for you.

Paul Bruffett (10:33)

Yeah, you gave me a lot of directions I could go there. I mean, what I'd say is exactly as you highlighted, right? mean, retail in general, I would argue has always been very competitive, right? And I think QSR is especially competitive and doubly so now. I mean, a lot of companies ourselves included, we just did earnings, right? McDonald's did earnings, Wendy's, Starbucks not too long ago. And you see it pretty consistent across the industry.

The folks are fighting to maintain same store sales, right? The growth story is really units and maybe unit volume, like average unit value. But, but a lot of folks are really fighting to kind of keep what they've got, if I'm being honest across QSR. And I think retail in general, Walmart didn't have, you know, amazing earnings either. It was pretty downbeat. and so I think that leaves us with a few different areas, right? I mean, value is a huge story.

A lot of, a lot of companies, ourselves included, are leading with value and a value story to folks. But how do we, how do we make that a reality? And I think as we think about the data, right, as we think about some of the things that you were asking about, it's really for us to bring together the information for our business users to really optimize the way the business works. And I think that's historically been a challenge. So I worked with, in my consulting time, I worked with a national grocer, and we built a supply chain control tower. 

So we brought in all of the data and they were massive. I mean, billions and billions of records, hundreds of millions of updates every day, because we're talking about on order, in transit, in a distribution center, shipped to the store, in a store, forecasted for all the products that a national grocer has, which is a huge assortment, right? And the challenge for them was they couldn't get that in one place, right? Somebody that's procuring sees in one dashboard, in one unit of measure. Somebody in DC sees a different dashboard, different units.

Somebody in the store uses even a different units, right? They couldn't even talk apples to apples to each other because they were talking about different units of measure for products. And so it made it really challenging for somebody to then go say, okay, I got some more Gatorade. I got 5,000 units of Gatorade. Where do I send it to have the biggest impact on my business? Right? They couldn't answer that question in a consistent way. It took a lot of phone calls and a lot of telephone game for them to get that. And don't even get me started on regulatory reporting. Cause that was whenever there was an infant crisis, right? And how much you send in to where and are you doing your due diligence? Like they struggled with that, right?

And so I bring that up to say a lot of companies are going through similar journeys in my experience, us included, of bringing together data into data products, right? For these different domains that are really the key legs of our business, right? Things like customer and store and product and transaction and supply chain and building these data products so that our business users can get integrated insights, like I was talking about, but also get insights across those domains and link them together to build an, to build a comprehensive story, right? Because ultimately what we're trying to do is maintain supply chain margins, right? The average unit volume and like costs to build stores, right? Integrated with customers and how those are going to perform cannibalization. It all really comes into a connected hole, and it still remains a challenge to be totally honest, to build sort of a global optimization model. You know, maybe that'll be a nice idea for somebody to build a product, but even optimizing within those and contextualizing across them continues to deliver a lot of value for our business.

And we still have a lot of work to do to finalize and develop those data products, but we've got the platform, we're making significant investments over the last several months and into the next couple of years to build those out and to democratize more of those insights like I was talking about on the back of the platforms and on the back of those data products.

Kelly Kohlleffel (14:01)

Yeah. Yeah, absolutely. Well, since you and I've known each other, I think when we first met, data was just starting to really move to the cloud. And obviously that's everything, you know, everything is in the cloud today. Do you want to explore a little bit, kind of that move to the cloud, where you are today, how that has shifted data stacks, what's going on there either in just retail in general or even if you want to comment on Jack in the Box and your overall brands and then anything from a developer standpoint I think would be really interesting for the audience.

Paul Bruffett (14:43)

Yeah, I mean, again, you gave me lot of territory to work with there, so I appreciate it. I think, you know, talking about data stacks, it has been interesting seeing this evolve, right? And so I think whenever we started chatting some time ago, Devon was one of the first consumers of Snowflake, if not the first, in Oklahoma, and it was a relatively new product at the time, right? A lot of...

It didn't even support single sign-on from Power BI, if I remember correctly, right? And Databricks and Snowflake were natural complements. Maybe still are, but the co-op petition is really intensified, right? You see both of them moving into each other's spaces and arguing that they're full, sweet solutions. You do data science on Snowflake, which I've done for various clients in my time and Lake House on Databricks, which to be clear, Starbucks does, right? 

They use Databricks extensively and have built really integrated solutions on that. And so they both offer a fully featured things, which has been interesting to watch. And then you have the hyperscalers, the Microsofts and the Googles coming with their own solutions. And so I think you almost get an embarrassment of riches building a data stack these days. I think it also leads to some interesting questions of area of focus and you know, making sure vendors are kind of keeping their eye on the ball of why you, why you brought them into the first place. didn't, you know, I didn't bring them in to do BI for example, but increasingly you can do that, and I can build web apps, and I can host containers. And so it's a, it's kind of an interesting, it leaves us with some interesting choices, but, you know, mean, the value is still tremendous, right? 

One of the technologies that I worked on before bringing in Snowflake was SAP HANA. And it's great for what it's there for, for the, for traditional data warehousing, I would argue, but we really struggled to bring in other kinds of data and to scale to the kinds of analysis that our business was increasingly asking us to do in oil and gas. And so, I mean, the core value proposition is still there. It's still tremendous. It's just interesting to see how this space is evolving and the level of competition going on between some of these key vendors and how that will shape the partnerships going forward. Although, and last thing I'll say is, one of the things that I feel like I heard the most about in addition to GenAI, out at the AWS conference was Iceberg and how it's starting to break down some of those walls, right? And now maybe I don't have to do as much choosing. Maybe I get this open table format that really allows me to unlock my data from some of these vendors that may have been a challenge in the past.

Kelly Kohlleffel (17:11)

So I'll just, I'll start on that. There was again, a lot of really cool stuff there we can, we can talk about, but you, mentioned Iceberg, you mentioned Open Table Formats. When you look at your data program today, what does that, what can that potentially bring to you that you don't have today or where can that help you extend? You talked about this value within the organization that your program needs to, constantly bring on a daily basis. What does this concept of open table formats give you at Jack in the Box?

Paul Bruffett (17:44)

I think, you know, for us it's, it, what I hope is that it gives us more flexibility for the right tool, for the right problem. Right. Because what's non really a non-starter. And again, it was one of the key reasons that Starbucks use Delta Lake so extensively, right. Is it can be both prohibitively expensive for some of these massive data sets, right? You think about all the stuff that you know about a customer, all the transactions prohibitively expensive to duplicate that data to say, I really needed another place maybe for another kind of access or analysis pattern. But also it leads to multiple versions of the truth. Right. And so my hope is open table formats will give us more flexibility to store the data one time to access at one time and literally keep one copy for data and analytics while using tools that are right sized maybe for cost, right? Some, some AWS tools like glue. Extremely cost competitive, right? And do a pretty good job. And other tools like Databricks continue to lead, I would argue, in data science and building and hosting machine learning models and really running the ML ops lifecycle. But maybe I don't want to use that for my data warehouse, right? And so I think it gives us a lot more flexibility for these tools without necessarily locking the data into one specific product and hopefully using the best of each in a way that really wasn't, I would argue very feasible before, because I think there were lot of trade-offs with previous formats, including Delta. But I'm pretty bullish on Iceberg and some of these other modern table formats.

Kelly Kohlleffel (19:09)

Yeah, as you start getting more of these options out there, the decision-making, you kind of think of that decision tree, like which way am I going to go? I feel like I agree with you. I feel like that Open Table formats, Iceberg and Delta, these have enabled us to advance so far with data lakes in general. We still got a long way to go, but compared to where we were say 10 years ago. But the options now, do you choose different storage platforms or storage types based on use cases or truly do you feel like in maybe five, six, seven years you're gonna say, okay, it's all my storage is in Iceberg and I'm just gonna plug in the compute that I want, as you said, right tool for the right use case or the right workload.

Paul Bruffett (20:08)

I mean, I would love to see, I would love to see it all unified in Iceberg. I hope that's the case. To be honest right now, that's not exactly what we're doing. We're being selective about those use cases and still evaluating and really trying to think about where it would add the most value for us to utilize the format and where it gives us the most flexibility. But we're still, you know, we use Snowflake. We've adopted Snowflake and storing the data natively in the platform still adds a lot of value to us and still tends to be kind of the solution that's the default, I guess, if you will, right? Because it makes it the easiest for us to manage, and we get a lot of added value still with fail safe and time travel that you can get out of Iceberg, but is integrated into the product in an easier way that we're used to consuming.

Kelly Kohlleffel (20:01)

All right, Paul, so as we've been talking, what's been going through my mind is this timeframe around decision making in the data space. Been around ERP a long time as you have, and when you make a decision for SAP or Oracle, that's going to stick with you a while. You're not going to swap those out every two or three years, right? Your ERP is your ERP, is your ERP for a long period of time. Is it the same with data tooling and platforms? Do you think about it the way you think about an ERP or is it a much shorter timeframe or something in between? What's been your experience, and where are you today on that?

Paul Bruffett (21:38)

I think it's inevitably somewhere in between, I would argue, because it's definitely not like an ERP, right? I mean, you know, I don't think maybe we'll be on the same modern data stack as we define it today in 10 years, but I think probably not. I think, you know, but data does have gravity, right? I mean, I've worked with a lot of partners, it's a very competitive space in the hyperscaler world, right? Whenever I was in consulting, there was a lot of competition to get the data lake because data has gravity and it kind of exerts this pull for other workloads to follow it. 

And so I say that to say, I think it's true for the decisions you make for housing it. We're talking about open table formats. And I hope that allows us more flexibility in how we move these, but even that, you know, once you pick a format, there'll be a lot of, it'll exert its own pull where you store those tools that you use to access it. The workloads you build on top of it become pretty sticky, I would argue. 

And so I think it's difficult to move them. I think it's becoming easier, especially as the technologies become more modular. We're not using them monolith cause like SAP HANA, whenever you use it, had to use HANA studio, and all the stuff that you coded was like proprietary in that. Now that's not the case, right? I've got a combination. I've got, I've got Fivetran, I've got Snowflake. I've got dbt, I've got Airflow, and I'm using all these tools together. And so it makes it a little bit more modular as I think about bringing other solutions in and how those best fit into my architecture. But I still think, you know, it's challenging to migrate solutions and to be totally honest, the business almost never comes to me and says, Paul, I would love to move from Redshift to Snowflake, or I'd love to move from Snowflake to Databricks. They really don't care about that. What they want is insights in a more timely manner. They want to be able to get granular data in their dashboards. They want to be able to do new things, build new data-driven products and services. And so I think that makes it really difficult to, you know, build a business justification for, for move migrating with any kind of regularity, because I'd argue it doesn't deliver inherent business value to do so. Right. I think it unlocks capabilities in a lot of cases that do drive business value. But it tends to be a big project, and I think it remains a big project. And it just doesn't, it doesn't make sense to do very frequently as a result.

Kelly Kohlleffel (23:55)

Yeah, I look at, say the last 10 year time, last decade where if we step back and I think, you you probably don't remember these or you probably remember these days about as fondly as I do, which is not that much like 2015, you know, deep in, in Hadoop land, right? And that was, that was pretty tough times. That's only been 10 years ago. And I think about the three...major Hadoop distributions at that time, Hortonworks, Cloudera, MapR. And then I looked to the two major independent platforms today, Snowflake and Databricks. Both of those companies are north of $3 billion annual revenue right now. So there's a $6 billion market just between those two companies. It literally dwarfs those three companies combined 10 years ago. And it kind of speaks to your point like data has gravity and it is just continuing to increase and the companies that do it well, the platforms that do it well. I feel like if they, if they treat customers really well and continue to improve their product and deliver that value, you know, you might see maybe it's those two, maybe it's somebody else turn into some really, really large companies in the next 10 years. Thoughts on that?

Paul Bruffett (25:12)

No, I absolutely agree with that. I think both of those, I've used both of those companies, both of the products from those companies for a number of years now. And I find that they continue, they continue to be leaders in the space. They continue to innovate. I think what the hyperscalers do very well is chewing up the ground behind companies, right? To take what is a differentiated offering and make it, you could say commodity, but really make it table stakes as they integrate it into their first party offerings.

But I think that's great. I think that's a very healthy balance because it forces those companies to continue to innovate. It forces Snowflake and Databricks to continue to innovate. And I think they've done that. Even though I argue, hey, I want them to continue to focus on what I would argue their core strengths are, it has a ton of value for me to be able to build lightweight web apps directly in Snowflake that allow my business users to update data. Because at Starbucks, we had to build a whole framework for that. We had a whole custom development solution. Because there's a lot of data you need to capture about your business. 

Sometimes you only need it for a little bit. Like during COVID, we had to work a lot to capture vaccination status. had to do that in a very agile way because it wasn't a data element we ever captured before. We didn't capture people's vaccination status. Are the restrooms open or closed at this Starbucks, right? All that information was important for our business and we had to do it very quickly, but we didn't even have a place to put it previously, right? And so now in Snowflake, I'd argue it's dramatically lowered the barrier to entry, to systematically capture data like that and to make it available and reporting. I think that's a great innovation, right? And so I think to your point, I'm bullish and I fully expect both of these companies to stick around. I think Databricks, could talk about all the innovation they've done too, right? I think MLflow continues to be a leader there, right? They moved into a whole new space that whenever I adopted the platform, they weren't in, right? And made it an open source project, but continue to deliver first party value by making it integral, right? To the way they track experiments that link back to notebooks, that link back to the outcomes that link back to the deployments, I think that's extremely compelling still, right? And so, yeah, I mean, I'm on the same page of, think, right now, both of those companies, I expect to continue to see them as leaders for a decade, I hope.

Kelly Kohlleffel (27:21)

You and I talked about a bit at re-invent to gen AI and it's interesting because, you know, certainly Databricks and Snowflake had both been around for quite a few years when genAI burst onto the scene. So you could, you could almost say that really neither one in that particular area of AI had, had a jump, had an advantage necessarily. Certainly Databricks has been more well-known AI/ML in general, but, the gen AI space.

And so do you feel like that also could be one of those catalysts? I personally don't think it is going anywhere anytime soon. I think we're just going to continue to grow that to see us again, get one of these companies, multiples of these companies, the hyperscalers as well, up into some very, very large organizations delivering value. GenAI is driver for that.

Paul Bruffett (28:14)

I do think so. I'll tell you, I've done a lot of work with the PurePlay folks in the genAI space. I think they've got great innovations. I think a lot of companies, like Out of the Box, like my own, are really going to see a lot of value from GenAI, maybe whisper it as a feature in products as opposed to a standalone offering. And so what I mean by that is like Tableau, working to integrate genAI. So you can ask questions of the dashboard because we're not going be able to make every variant of every dashboard for every question somebody has. But a lot of our stakeholders don't want to or can't write their own queries. There's a lot of domain knowledge that goes in that. There's a lot of technical knowledge that goes in that. Right. I think genAI has a huge amount of value there. I think it also has a lot of value in the data layer, right. To help the business analysts. Cause we were talking about democratization, right. And I'd argue democratization is a continuum.

I've got folks that want to do data science. They just want an environment to write spark code or, you know, and, and go after it, right? All the way to these people that just want to get a dashboard email, but they still have questions about it. Right? We still get a question of why are my sales the way they are in California? Right. Or, hey, are we comping over a promo from last year? Or what is, is this year over year for Thanksgiving? Is it on a different day? Right? All these things generate questions for my business users that they inevitably email somebody to ask, and then that person clarifies. And I think genAI really soon will be able to at least take a first swing at that and give people context about the data they're seeing in a meaningful way that'll help them kind of get to that own, get to the decision, get to the insight, get to the outcome that they really use the dashboard for. And kind of to your point, I do think that we'll continue to see that's the reporting layer, but I think there's still a huge opportunity in the data layer, right? 

To say, really, I want to empower folks to just go ask questions of the data and get insights in a much less structured way, if at all possible, and to really build more of a published layer of data products that people can go interrogate and do in an open-ended way with maybe less work required from the data team beforehand to structure the question, you know, and contextualize it with maybe our strategy documents, maybe some of our competitive intelligence information, it's not structured, right? It's unstructured data that can augment or compliment the insights that they'd be able to derive or ask of the structured data that is resonant in those platforms.

Kelly Kohlleffel (30:41)

Are you seeing value, Paul, for genAI for both structured, certainly unstructured, but structured data as well? Use cases where both of those data types apply.

Paul Bruffett (30:51)

I definitely think so. I think to your point, a lot of the use cases that I initially saw, especially in retail were around unstructured data. I mean, one of the first, Accenture put it in Accenture.com, right? If you go ask a question on Accenture.com of what was, you know, compare fiscal year 23 to fiscal year 22, genAI I'll take a first swing at that using all of the hundreds of thousands of documents at its disposal. I think that's a great use case. We also work to augment search even just in a generic way, right? Because I still go to websites and don't get tremendously relevant results for some types of products. GenAI can actually help behind the scenes, even if you don't change the search experience. 

So I think some of those to your point were some of the more obvious quick wins, as well as improving product data quality. I actually saw a number of retailers doing that, but structured data has been a little bit slower and the persistent challenges or questions that I saw was how do I trust the answer? Because with unstructured data, like you go ask Google Gemini a question, and it'll cite the answer, right? It'll say, “Here's the answer, and here's the article I found it.” You can go check its work. With structured data, it's almost paradoxically more difficult in that I could show you the query, but are most people going to be in a position to validate the logic for a genAI generated query? Probably not. Then if they, so how, what's the risk worth to folks that they got the wrong answer from genAI to their business question on structured data versus going to somebody. And if somebody from the data team has to check the answer anyway, why didn't you just start with them? And so I do think that continues to be a little bit of a stumbling block, if I'm honest. And that's why I think there's still a lot being worked out from my perspective at least, but back to my previous answer, I'm still bullish on the technology. I just think there's, we've got to figure out some of the explainability portion of it a little bit more thoroughly at this point.

Kelly Kohlleffel (32:39)

I love that. Last thing here I want to cover with you. I always love looking out and I'm not very, I'm terrible at it. Actually. I'm not a crystal ball guy. I kind of look at, know, what just happened? Ooh, yeah, that was a great idea. But what do you see? What are some of the trends you see? Retail. Food and beverage industry and how can data leaders in retail and in food and beverage but across the board really start preparing for these trends that you're seeing, these emerging trends in the coming year? What should they be doing?

Paul Bruffett (33:11)

Yeah, I mean, maybe similarly because my crystal ball is pretty foggy. I see it as continuation of existing trends, right? And so we've got a lot of cost pressures and so we see a lot of automation, right? Kiosks are gonna continue to get deployed, but I think that's really exciting because it provides more opportunity for personalization, right? We found that people tend to spend a little bit more at kiosks, probably because they get longer to browse, because they get more interaction with the menu, right?

And I think that's, I think that's exciting, and it's opportunity, but it is evidence of automation, right? Increased automation, which again, I think is great. And it's going to allow folks in the restaurants, in the stores to focus on really interacting with guests, getting food out in a timely manner and maintaining that customer experience, that guest experience and redeploying labor. And similarly, I think it's supply chain, right? Is cost pressures across that and really trying to contain that. I think we continue to make that a focus area. I think a lot of our peers are doing the same thing.

And so I think we'll see work there, but really I'm, I'm excited about continuing to see some of the new technologies. I mean, there's still so much opportunity for deploying decision intelligence with, with new machine learning models in new domains for us. And I think for a lot of QSRs, if I'm being honest. And so I'm really excited to see some of the new opportunities for that, especially how we personalize our experience more comprehensively because like a lot of QSRs, historically we didn't have that much data about our customers, right? Loyalty programs, maybe McDonald's hasn't been doing theirs that long. We haven't been doing ours that long. Starbucks was obviously a first mover in the space for QSR and has been doing it for a while. I think there's a lot that we can do, and you see them reinventing themselves, right? And so I think it'll be interesting to see how these businesses that have historically been very physically focused, right? We're driving towards 20% of our sales being digital.

How does that change the way that we think about the business? How does that change the QSR industry as we become an increasingly digital business that serves food out of restaurants, but gets interacted with in a number of different mechanisms, a number of different channels and try to become much more pervasive and personalized.

Kelly Kohlleffel (35:17)

Everything you're talking about, this digital interaction, personalization, supply chain, all that driven by data too, right? Awesome. And Paul, this has been fantastic. Really, really have enjoyed catching up with you. We got to explore some fun topics today. Thank you. Thank you. Thank you for joining the show today. 

Paul Bruffett (35:34)

For sure. Thank you for having me.

Kelly Kohlleffel (35:35)

Absolutely. Our pleasure. Look forward to keeping up with everything you're doing at Jack in the Box as well. Thanks to everyone. Huge thank you to everyone who listened in. We really, really appreciate each one of you. 

We'd encourage you to subscribe to the podcast on any of the major platforms, Spotify, Apple, Google. Find us on YouTube as well, Fivetran.com/podcast is where you can visit us and please send us any feedback or comments at podcast@fivetran.com. We'd love to hear from you. See you soon. Take care.