Inside SEP: What a Real Data Strategy Looks Like in Software Development

Jason LaJeunesse 00:00:00

Thought about this in kind of two elements that I think are pretty critical. Part one is just we need to realize, you know, large language models, they're not human. They don't have history of your business or context.

They may not understand what's safe to do and what's not to do and why. And it's not always, you know, clear.

So having good grounding on your business with your data, as well as being able to look up things safely is really important to be able to enable them to work effectively and safely.

Zac Darnell 00:00:38

Welcome to behind the Product, a podcast by SEP where we believe it takes more than a great idea to make a great product.

We've been around for over 30 years building software that matters more and we've set out to explore the people, practices and philosophies to try and capture what's behind great software products. So join us on this journey of conversation with the folks that bring ideas to life. Hey everybody, welcome back to the show.

I am your host Zach Darnell. Got two guests on the show today, both from scp.

I asked our new director of data, Jason Lodgeness and one of our solution architects, Jason Dossett, to join me to share a little bit about how they and SEP think about data.

We talked through how SEP thinks and defines data, why we think it's becoming more and more important for companies and some of the shifts that we've seen along with what we think is important for really any business taking on data problems. Jason and Jason share their unique perspectives and what they've seen work well in real world scenarios.

And yes, we do touch on AI because there is an important connection between data and AI and I'll define AI in the context of LLMs and generative.

So if you've been following along this data series, I think this will be another lens to help you better understand how you could navigate the world of data. Thanks so much. We'll dive in. So Jason La Genis joined us here in December right at the tail end of the year as our new Director of data.

Relatively new practice. Not that data is new to sep, but it as a service line, it as a practice, it as a capability in the building new worked on this a lot last year.

Jason Dossett, solution architect here, myself, our CEO Raman, our Director of Engineering John Fuller and a few other stakeholders across SEP really spent about 12 months validating the market, testing some opportunities, looking at and exploring whether or not this should be a thing that we offer at SCP and just thought it'd be a great time now that Jason Lanis has been with us for a few months. Let's talk about where we're at, how we think about it, what we've seen, maybe where we think things are headed.

So that's kind of the couching for the conversation today. I just thought that would be interesting.

So as we think about the smattering of things I just talked about, I think it'd be good to start with defining the thing, like what do we mean when we say data?

Because I think, know, Jason Dawson, you know, when we were even talking about that internally and, and doing updates to other folks within SEP last year, like even saying, here's what we mean when we say data, that took some iterating. That was, that was, that's a hard thing to communicate because there's a lot of nuance inside of that. So I'll just toss this out to both of you.

Like when I say data as a service or as a product for scp, what do we mean when we say that?

Jason Dossett 00:03:45

I can start with historically, what it doesn't mean, which is important. I think, I think it's really good.

Yeah, we're not talking about operational data store supporting applications traditionally the way we built them as custom software apps, which.

Zac Darnell 00:04:03

Is, which is basically our business model for the majority of our existence.

Jason Dossett 00:04:08

Yeah.

Zac Darnell 00:04:08

Yeah, that makes sense. Okay, how would you define what it actually is then? I do think that's great. That's like a great other end of the spectrum. What's the other side?

Jason LaJeunesse 00:04:20

I think what we're seeing is that data is becoming much more strategic.

And so obviously we've needed some elements of data components in those individual applications, but we're kind of seeing in modern data platforms becoming a much more strategic part of your business.

Not just do I have an individual ERP or CRM system working or an IoT data off on its own, but what we're seeing is I don't just need to organize that data centrally to be able to do some BI analytics. It's not just the analyst that cares about that data warehouse anymore.

We're seeing a lot more downstream use cases of those modern data platforms and more reasons to kind of get that centrally stored. And as we'll talk about a bit, AI does change the game a little bit.

Jason Dossett 00:05:03

Sure.

Jason LaJeunesse 00:05:04

Obviously these trends have been happening for a long time, but it kind of adds some fuel to the fire as far as like, why is it important now?

Zac Darnell 00:05:10

Hmm. You know, that's, that's actually interesting. So, you know, I know we'll talk a little bit about like, you know, why we think this matters now.

But it, like it wasn't just AI like the, this data as a product has been around longer than LLMs and generative AI has been around. So it's like maybe it was gas on a fire that already existed.

But what do you think if you look back, maybe, I don't know, five, ten years, what were, what was happening to where customers were thinking about data more comprehensively?

I don't know if that's the right word because it's more than just in service of a single application or more than just in service of a single power BI report. It was more around enablement. I don't know. Could you talk to any of that?

Jason Dossett 00:05:52

Well, I would say, I don't know that there's been a bigger, I say a bigger focus. It's just more on our radar because the platforms are now more aligned with how we've built software.

People have been trying to put their data in a single place and make it make sense for a long time.

So between the platforms and some of the better patterns that have come around, like Lakehouse medallion architectures that have kind of figured out some best practice and better ways to do things from, hey, let's try to build the perfect schema of our entire enterprise and map everything to that schema. Like that's what people were trying to do for a few decades.

Jason LaJeunesse 00:06:41

Yeah. And I think to kind of going alongside that historically go back even further.

A lot of these workloads were very focused on a DBA working through what's the perfect schema, how can I get the state in one spot and kind of the aspect of uprising of things like data ops, analytics engineering, that's really been kind of a trend that has worked well with SEP because we're seeing more and more folks taking a software engineering focus. How do I take things like DevOps and apply them to the discipline of data warehousing or data platform building.

And so that's been, you know, I think a big reason why SEP has started to kind of say like, hey, this is actually a good fit for us because that really fits into our background of, you know, strong software engineering.

So analytics engineering is looking at, you know, how do I take things like a more version controlled approach where I am using best practices with DevOps now we have the terms like data ops.

How do we build a platform where we can actually, you know, build in all of that reliability and resilience using those skill sets we learned from the software engineering world.

Zac Darnell 00:07:48

That's really interesting. So, okay, I want to part like there's, there's a lot of terms, roles, discipline areas that you just listed off.

But just think about some of the big ones that I feel like I've been having a lot of conversations around. Data strategy, data architecture, data engineering, platform engineering.

How do we think about some of maybe the big four or five key roles around data as a product, data as a more integral piece to somebody's business? Because I think I hear a lot of different versions of what different people mean when they say those same words.

So in this world, I don't know, you pick maybe the top three or four that I just listed off. How would we define those things to somebody else outside this building?

Jason Dossett 00:08:35

There's definitely a blurring, I think the differences like focus being on the data. So data engineering, you're processing data, you're not necessarily building a bunch of business logic like we would do as a software engineer.

And the tools use data Architect, more focused on structure of the data, how we're processing the data, as opposed to someone that's more concerned with the components and the platform to build the software that's going to provide the data architect the tools to accomplish those things. Yeah.

Jason LaJeunesse 00:09:14

And then just to round out, analytics engineering is a newer term in the space, probably the last four or five years, but seeing a lot of progression.

But the idea there is, hey, in the past maybe you would just build a one off SQL Query or a one off View, that's just something that's stored on your hard drive or some shared team drive. But it's like, hey, how do we actually pull in tools to use best practices from engineering?

How do we get that into something like GitHub, how do we deploy that in a very controlled way and how do we have things like automatic documentation that makes it very clear to understand, you know, how do we get access to that data, how does it integrate with other systems? So you know, that's where we're kind of seeing.

It does have more of an analytics flavor because it is focused on SQL, which is a little bit different in that it's declarative.

So I think maybe something that's a little bit different than a lot of our traditional software engineering work, which is maybe sometimes declarative, but you know, typically more procedural. So it's kind of a different flavor, but has a lot of things that resonate from the work we've done in the past.

Zac Darnell 00:10:16

Okay, so thinking about some of the work that we've done in the past, who we've been to, where we are now, okay, we mentioned a little bit ago data as a product and the maturing of data as a service line is happening long before Generative really hit the market a few years ago. But AI being that gas, that might have made this more apparent to folks. Why do you think that is? Like why, why was AI?

And what I mean by that is not the complete umbrella of AI specifically mean, you know, LLMs and generative. Why was that the gas that, that made data way more top of mind for folks and way more strategic.

Jason LaJeunesse 00:10:55

I actually think I've thought about this in kind of two elements that I think are pretty critical. Part one is just we need to realize, you know, large language models, they're not human. They don't have history or context.

They may not understand what's safe to do and what's not to do and why, and it's not always, you know, clear.

So having good grounding on your business with your data, as well as being able to, you know, look up things safely, is really important to be able to enable them to work effectively and safely. And so that's part of it.

And the other thing, kind of the other side of it is from a opportunistic perspective and also thinking in light of the fact that AI is commoditizing some things so like writing a hundred line SQL query, you know, a lot of these agents can do just fine on their own if you give them enough context.

There's kind of the question of like, you know, what is valuable to our organization and something that can be uniquely valuable to you is your proprietary data.

And so there may be some things that are, you know, obvious out in the market that agents can just pick up and do because they're able to learn on all that data that's out there.

So now there's kind of this new potential proprietary data source though that, that could be a new value for your business that maybe wasn't as apparent before.

Zac Darnell 00:12:11

That's interesting. I haven't really thought about the, the, the first angle that you took. Like that's actually really interesting to me.

Jason Dossett 00:12:18

Yeah, I would say there's some things on the ground that have kind of impacted AI, like just reinforcing the fact that people didn't have their data in a good place and everybody's been trying to build the perfect warehouse, lake house, whatever, for a while and you can get by with not having one because you get smart people and they know how to find the data and query it. That no longer works as well when you're trying to do AI.

It actually needs to be well structured and clean and all the things that a person can't just do ad hoc just in time. Like it needs to be there.

So it's like, well, we can't just get by with kind of the hero aspect of people having smart people be able to find the data and create reports ad hoc and do that hot querying. Like, no, the data needs to be in a decent place to at least even get started with AI.

So I think that's why it's really, it's like shined a spotlight on all the nooks and crannies of like, yeah, our data isn't really ready. Where in the past it was like, yeah, we could get by without it being in a good place.

Zac Darnell 00:13:33

I think that's, that's actually, that's another really good point.

I think that's one thing that we've heard from even some of our, our customers over the last year where they've tried to enable, even if it's just, you know, I don't know, a chat bot inside of their company or, or co pilot or whatever and realize, oh no, so and so shouldn't be able to see that information or you know, I tried to generate this report and that's all wrong and hallucinated all over the place. It's not just as easy as turning it on.

You really do have to kind of get to some basic foundational levels before you can enable this stuff really at scale across any business, you know, is there, I would imagine that there's some version of air quoting like data work that is table stakes, that's, it's foundational, it's plumbing, probably not, you know, super snazzy, but very necessary.

How do you draw a distinction between that, Maybe not even a distinction, a connection between that kind of work and the strategic nature of data being part of your strategy. Like how should customers be thinking about this or really any company be thinking about this?

Because I again, I could see somebody thinking, well, you know, that, that's just, we just, we just need to set up a more mature data translation pipeline. It's like, yeah, that might get you part of the way there, but there's more to it. And here's why.

Jason Dossett 00:14:58

I think there's an understanding of like what you're trying to accomplish. Yeah, that was a, there's a loaded question there.

Zac Darnell 00:15:05

I know, I know. I don't know. Take it, take, take any piece of it or we can throw it away too.

Jason LaJeunesse 00:15:10

Yeah.

And I think it, a lot of it does get back to the fact that again, agents don't have the context to know, especially from a security or, you know, safety perspective.

They don't always know how to make the right decision that a human would just have the common sense to do of like, can I send this piece of information to this other person via email? That may be very obvious to many people in your organization for different contexts, but it is very common for an AI model to struggle with that.

So I think a lot of it is like, hey, you know, there is a lot of potential value to be gained by, you know, automating a lot of these things with these agentic systems. But you can really only do that if you have all the systems in place to enable you to do that safely.

So things like cataloging your data, classifying it, making it clear you have role based access controls that make sense, maybe you need to build some MCP servers that those agents interact with to make sure that they're only able to see the right things. You know, kind of getting all those basics in place is what's going to enable you to be able to use those agentic systems effectively and safely.

Jason Dossett 00:16:17

It's a good point around like the governance, which is again kind of hearkening back to why now, like in those areas where people were just doing the queries and figuring out and there was, it was humans doing that, like you could get by again without having the governance as much as without the structure. Whereas now, yeah, like Jason just mentioned, like the agents aren't.

You can't just replace that like superhero with an agent because the superhero knew where the data was and how to protect it and what made sense, where the agent needs to know, which means you need governance in place. The hardback. Yeah.

Zac Darnell 00:16:59

Do you think that like should companies be thinking about governance first or should they be thinking about systems first? Is this a holistic view? Like how should, how should somebody, if they're.

I'll paint just a quick picture because I've actually heard this from a couple of folks. Hey, we tried to turn on something again, insert, pick, pick your harness, pick your, your, your tool.

And we didn't either we didn't get the gains that we thought we were going to get. It wasn't as easy to enable.

We realized very quickly that people had access to the information they shouldn't have access to or not access to the things that they should. And we need to go back to the starting block and really rethink our strategy for enablement.

Or what should these folks, how should they be thinking differently? Like what's the shift that we would tell somebody if they want to head down that, that path.

Jason Dossett 00:17:53

Yeah, I think there's a balance. And being more goal oriented maybe where in the past a lot of these data efforts were kind of attempted to be big bang.

Instead of trying to define your entire governance up front and then implement a big bang towards that being more goal oriented to where you can incrementally get your data better to solve the important problems that you're trying to solve early. And that can help inform, like it's just a cycle. Can help inform what your governance should look like because you can identify where the risks are.

Zac Darnell 00:18:26

Okay.

Jason Dossett 00:18:27

That kind of thing.

Zac Darnell 00:18:28

So you don't feel like you have to have it all understood up front. There's. There is, yes, you should have a vision, be centered on the problems and the pains that you have. That, that is good.

But there is incremental steps in growth maturity even within your governance strategy that can be had over time. That's okay.

Jason Dossett 00:18:44

Yeah, I, I think so. I think you need, well, strategy. I think you need. There's best practices and those kind of things. Have those in place that guide decisions.

But I think it can be more incremental. I think you'll get value quicker that way.

Zac Darnell 00:19:01

I think that's interesting. Sorry, Jason, I'll come back to you.

Like, I've, I've actually, I mean, you and I talked to some of our customers, I've, you know, talked to friends, you know, over coffee about. It feels like we can't find a way to, to, to build a strategy that is incremental value over time.

Like, it feels like we have to rip out the whole thing like that. That concept in building out a real plan and executing against that plan seems to be a hurdle for, for folks that I've talked to.

I don't know if there's a, you know, magic sauce or if it's just so contextually specific that, you know, there's no, there's no single way to do that. But it, that seems to be a very hard or big hurdle for a lot of folks.

Jason LaJeunesse 00:19:45

And I do think things like bringing an MVP mindset to how we approach it, you know, concepts like vertical slicing, I think that's exactly what we're talking to. Especially when you're working with something very new. So say you're new to working with agentic AI systems.

I think that can be really helpful to kind of say, like, we don't have to figure out the breadth of everything, but let's figure out end to end, how can we make this secure system with the data that we need to accomplish this outcome that's important for our business, kind of start there and focus on it.

And so I know like sometimes it sounds big to say, hey, we're going to roll out a data platform that sounds like, oh, that's got to be this huge massive super wide platform.

But you know, in the age of cloud based computing can actually be pretty quick sometimes to spin up a new, you know, data platform, you know, be that fabric or databricks or whatever, especially say in like a dev environment where you're testing this out.

So it is possible to definitely look at, you know, what is our end goal, what is like a good business use case and also have in mind kind of what is the strategic things we want to learn about say agentic AI and our data platform and how we approach classification and kind of focus on that vertical slice and work it through and then kind of build that out over time once you can find success on that use case.

Zac Darnell 00:21:02

So can we talk about, I love where you both are going with this. Can we give maybe an example from maybe a recent project? I'm thinking maybe aerospace client that we work with.

We've been working with them for what, the last year, year and a half on a specific data platform enablement piece.

And I mean, Jason Dawson, you, you kind of like came up with the initial strategy for this and I think took a vertical slicing approach to try to find incremental value over time. You maybe give an example for somebody if they were going to create a mental model for their own problem. I don't know.

Is that, is that even, is that a fair question?

Jason Dossett 00:21:45

Yeah, that one, that one's maybe not, not even the best example. Like I think the work we're doing for the energy.

Zac Darnell 00:21:54

Oh, okay.

Jason Dossett 00:21:55

And the energy sector is a better example because we're working kind of as much top down as bottom up where we have like a very specific use case. They need a report that shows this, they're trying to accomplish some insight based goal and you can work backwards from that to identify.

Well, okay, to do that we need this data to be cleaned and staged. Well, that means that we can trace that back to the source systems and source tables that we need to get that data ingested to support that staging.

So that becomes kind of an incremental approach as opposed to starting from the other side and saying let's take all our source systems and get all our source tables and dump it all in the raw. You can do kind of a top down, bottom up but the top down is based on goals, goal oriented increments towards that.

They already had a really good platform in place for adding new source tables, for example. So you build that platform, you get that ready for that incremental process and.

Zac Darnell 00:23:07

Then work your way through goal at a time. Yeah, okay, that's so you, so you could do both but like be anchored in a goal. That's what I'm hearing in that. Okay, that's helpful.

I think, you know, the, every context is different but there's got to be, you know, there's, there's probably some common strategies that folks could, could peel out and try in their own environment. So I think, I think that that is a great example of one of those. I want to, I want to pivot a little bit to like what are the teams and ownership.

We talk about governance, we've talked about, you know, analytics, engineering. That definitely that's a new one for me. Haven't heard that one before today.

What do you think might be missing inside of organizations today that they might need to think about from a role perspective or a capability perspective? Is it modern data thinking? Is it experience in specific technologies like Python or programming languages like Python?

Is it like what do you think that looks like for most companies? And I get, you know, I'm asking for a very broad swath here and I'll say and. Or what is pivotal to be successful, to execute kind of an example that.

Jason Dossett 00:24:24

You just gave something that's, I think you asked for, like what should they. Are there in terms of gaps or what should they.

I think something we've seen, maybe Jason can give a broader answer, but something like very specific we've seen is the lack of an enablement team. So you've got a platform team that maybe is building the platform and you have dev teams that are trying to build on the platform.

But then you also have your business users that are trying to use the platform and there's this gap of enablement where the platform's great and I think they kind of have the field of dreams mindset. Like if we build it, they will come, but it's like you need to bring them along.

I think seeing this in another project, like this concept of an expert workspace or whatever that the users can use, you need that enablement. Like it's not just a platform and throw people at it kind of thing. Seen that in several of our projects.

Zac Darnell 00:25:29

And this is matching between the platform team and the business users business, even.

Jason Dossett 00:25:34

The development teams, like being able to enable development teams to build Solutions on top of it. There's a gap there that.

Jason LaJeunesse 00:25:44

Yeah, and I do think that's where I remember who coined it. But software is eating the world and now it kind of feels like AI is eating the world.

Technology is moving very fast, oftentimes a lot faster than people can keep up with. So there is a sort of cognitive load problem right now across the board.

So I think even within the software engineering space, we are doing what we can to keep up with how quickly some of these AI tools are evolving. So surely if that's not your only day to day focus, this space is evolving very fast on the technology side right now.

So I do think a lot of what we are trying to think from a platform perspective is how do we abstract out some of those cognitive concerns. But I think there is an element of, there is a challenge where you can get this gap like that Jason was talking about of technology is moving fast.

We are trying to adapt to it as a technology company and as a maybe a technology organization within your company. But pulling along the business users into these newer ways of working and these newer tools is certainly a big challenge.

So I think that is where alignment top down across the organizations. So that's great if your CTO or CIO is really excited about some project.

But you also need your CEO and your CFO to understand like why that's valuable and why they need to drive that down into other parts of your organization as well.

Zac Darnell 00:27:11

Why do you think, or I'm going to ask one more question. What do you think that enablement needs to look like?

I mean we're really, we're not talking necessarily like the title in the building is probably not, you know, Chief Enablement Officer. That's not what we're talking about. You're really talking about a function or a role.

What is the enablement group or person team role doing to help bridge that gap in kind of the, in the space that you're talking about?

Jason Dossett 00:27:38

You took my answer, which is bridge the gap. That's the big part of it is.

Zac Darnell 00:27:43

Is it somebody that like understands the technology, that like Jason, Jason Longines, like what you're saying as far as like how things are moving and evolving but also understands the pains and the constraints of the organization. Like it's really somebody that can blend into both worlds?

Jason Dossett 00:28:00

Yeah, I mean there's an aspect of these platforms where it's like you can deploy it, but that doesn't necessarily mean it's like user ready. And so it's those gaps to some extent making it ready and purpose built for what you're trying to accomplish.

And that's why that enable is that gap is because the platform team's like, oh, the platform, you know, we built the platform, it works. People can go do whatever they want.

Well, there's also being able to do whatever you want is limiting because you don't necessarily know how to or where to use.

Zac Darnell 00:28:36

It's almost overwhelming.

Jason LaJeunesse 00:28:37

And I know there are some organizations looking at things like forward deployed engineers from an agentic AI perspective. Really this concept of maybe sometimes with hub and spoke models where you're embedding people into the organization, being very collaborative.

Because I know something we're looking at is if you're a platform team and you're doing internal user research and maybe you're talking to one of the potential users of your platform once every couple weeks for an hour or something, that may not cut it. It may need to be more.

You are embedded next to that person, collaborating very closely as you start trying to build out and roll out these platforms. Again, with the pace of technology changing, I think it's just incredibly important to be very intentional about how you do that.

So it could be a mix of more standard training for your users, but it could also be actually embedding people into those departments and being more collaborative.

Zac Darnell 00:29:28

I mean, this is probably an oversimplification. So I apologize and keep me honest.

It almost sounds like insert buzzword Marty Kagan embedded product team with a customer and just applying that to more data centric problems. Is that, is that a, am I, am I painting with too broad of a brush when I say that. And I, and I get that, like that's not an ex.

An exact quote from Marty Cake. And I'm just kind of like trying to throw kind of a mental model inside of what you're talking about.

When I hear that collaboration, it's almost like I'm going to pair a designer and an engineer together. Like that's an example of high collaboration that has existed in SaaS and software development development for a while.

This almost sounds like pairing some kind of, maybe it's an agentic engineer and business user to solve a business pain or problem. Like, I don't know. Is that an apt example?

Jason LaJeunesse 00:30:30

Yeah. And I think, you know, obviously patterns from the past in terms of how people work together or, you know, it's not going to be novel completely.

Like, I think applying that pattern is correct. I think that is essentially, if you hear the term forward deployed engineer, that is essentially looking at folks that understand AI really well.

That can understand business context well and kind of bridge that gap.

Zac Darnell 00:30:53

Got it.

Jason LaJeunesse 00:30:54

To make solutions quickly. And the reason is, again, AI is moving so fast, you kind of need people to be able to bridge that gap.

Zac Darnell 00:31:01

That's fair. And I know we're not talking about AI in the context.

This conversation was not to be focused on AI, but I feel like at this point in time, data and AI, they seem to be. It's impossible, it seems like, to tease those two things apart in any conversation. But it is more than just AI enablement.

Like, I'm also hearing it could be visualization, it could be software solutions, it could be whatever, but it is bridging that gap from a data perspective to those different business problems that are going on.

Jason LaJeunesse 00:31:36

Yeah. And that's absolutely correct. And it's worth noting. Yeah. Again, AI is kind of eating the world. It's hard to move away from it in any discussion.

Zac Darnell 00:31:44

I think. Maybe, I don't know.

Jason LaJeunesse 00:31:45

But we have seen as well just general use of the platform, getting used to ideas of using a data catalog or the importance of curating your data catalog and kind of working with departments to think through things like, hey, how do we define owners if he's going to look over this data set?

So there's definitely a lot of opportunity for coaching and to take that embedded approach for some of these other just like, best practices in managing these systems.

Zac Darnell 00:32:12

I think that's helpful characterization for me because I think I walked into this conversation thinking like, okay, well, if a client is stuck, if any organization is stuck in moving AI or data centric problems forward, it's got to be either a skill set gap or an organizational gap or something. And it's. There's no silver bullet. It really is probably a smattering of both and it's unique to every organization.

And it really just depends on Jason Dossett. Back to your point of what problem are you trying to solve? Stay anchored in that.

And, and that can be like the overarching strategy for that incremental value that, that maturity model over time. So I think I was looking for a silver bullet and there isn't one. And that's okay. It's just the way that it is.

So I want to, I want to maybe kind of move us for. Move us into kind of where, where we're thinking in the future, where we see this moving. Everybody has their own version of a crystal ball right now.

I think anybody that's predicting the future is making it up at this point because we don't really know. And to your point, AI is eating the world in a lot of ways. I am personally not in the.

Everything is going to go away in our world of technology and AI is going to take over every job. I'm not in that camp personally. Maybe that's because I'm optimistic. Maybe it's blissful ignorance.

We'll find out what are things that companies are not aware of yet that they need to be thinking about when it comes to data as a strategy. That's really what I'm getting at in that question. I think I'll give an aha some of the ahas. Oh, I can't just turn on AI and it's just magic.

Like, okay, that's an aha moment people had to come to, maybe through a little bit of pain and cost and time. What are things that people haven't had the aha moment to that you're. That you.

That either you two see or have heard when it comes to all this, this world of data.

Jason LaJeunesse 00:34:11

If I just think about where things are going, I mean, I think a lot of companies are starting to realize the importance of focusing in some different areas. But something that we've thrown around a lot in this talk is the word platform.

And just something to keep in mind is I do think platform engineering as a discipline is going to become a lot more important. And so it's just something to be aware of, especially with AI commoditizing some of the things that happen below the platform.

And there's this idea of working on the harness instead of in the harness.

It very much mirrors, I think too, what we're seeing in the data space to say, like, we need to make sure this data set is very curated, it's easy to understand, it's well controlled, we know what can be done with it. There's new methods coming out all the time for better ways to take advantage of your data.

I think there's going to be a lot more focus on how do we manage those platforms. So it's just like a skill set and discipline that I think organizations should consider starting to spend more time in.

Zac Darnell 00:35:14

Okay, yeah.

It seems like some like, you know, you two have even helped educate me on, you know, I think I was drawing a very hard distinction between platform engineering and data engineering. And while that may have been true, it doesn't. It's not always true.

I think is is the new conclusion that I've come to that, at least for us, that can be the same, that can be the same person on a team and there's maybe some nuance to the way that that person shows up in the context of a data team, I don't know. Is that fair?

Jason Dossett 00:35:46

Yeah, I don't know if it's super blended. Like, I think the platform engineering, you're providing the tools for the data engineers to build the processing and those kind of things.

I think maybe there's multiple tiers when they can fluctuate whether platform engineering is getting closer to data engineering or vice versa. Vice versa.

Zac Darnell 00:36:12

That's fair. Maybe it's. Maybe it's depending on the organization. Maybe they're less distinct.

Jason Dossett 00:36:17

Yeah, yeah. I think a lot of it's based on your business need, based on how your organizational structures already set up.

Like, if it's somewhat stratified, you don't want to just, oh, well, we need to blend these because someone said they should be more blended.

Zac Darnell 00:36:35

Got it.

Jason Dossett 00:36:36

You want to start where you're at and work towards what makes sense.

Zac Darnell 00:36:41

No, that's helpful clarification for me.

Jason LaJeunesse 00:36:43

Yeah, I think part of the general trend there too. Just mentioned, like platform engineering.

I know we've been working on things like more declarative ways of working, really kind of gets to the fact that information is moving fast, agents move fast. You can have lots of them running in parallel.

So it's more important than ever to make sure you have these platforms designed where you have all of your governance in check, where you are reducing the cognitive load of the people and the agents working in the system just because the space is moving so fast. And if you don't have those controls in place, you know, it can be very hard to manage all that.

Zac Darnell 00:37:16

Okay, so I'll throw kind of maybe like a closing question out to both of you. So if anybody listening to this, right, they're. They're their cto, their cdo, whatever, and they're in maybe a larger organization.

What's the, what's the one thing that you think folks across technology, product data need to be thinking about, need to walk away from? Is there a differentiated SEP position? How do we think about this differently, that we think people need to really understand or adopt?

Jason Dossett 00:37:51

Yeah, I don't know if this is necessarily even data specific, and I think it's kind of related to how people are slowly learning, like, oh, our data's not ready enough to also not thank the AI. And I think Jason touched on a little bit of this too, is not to think the AI is just going to solve your problem.

If you just wait long enough, the AI will be smart enough to just figure it out. You need to create those context boundaries for the cognitive load for determinism.

Because the AI is not always, you know, not typically deterministic. You need to create the boxes where at least inside the box, maybe it's not, but coming out of the box it needs to be.

So just not assuming AI is going to just figure that all out, like identify where those box boundaries should be, what the interfaces between them are, so that you can have determinism kind of incremental.

Jason LaJeunesse 00:38:48

Yeah.

And it, I think as far as, you know, closing like what, what's maybe different now from five or ten years ago, I think it's just noting that like we talked about before, you could get away a lot more in the past when it comes to not having some of these governance things in place, not having your data well curated in the future, you're not going to be able to get away with that. And so it's just noting, you know, we see it as increasingly important to do all these things that Jason Dassa was just talking about.

And so in the future, to take advantage of AI to have a successful business, you're going to have to have a strong data platform to be, you know, effective in the market.

Zac Darnell 00:39:28

Thank you very much. Boys, Gents. Boys. I can't say boys, that's unprofessional. We'll say thank you very much, both. Jason Law Genus. Did I finally get that right?

I didn't get it right, did I? La Genis. Bam. Only took 14 million tries. Mr. Daw said thank you so much.

Jason Dossett 00:39:46

It's actually D. Shut up.