Leadership Insights: Data Observability in the AI Era with Pantomath’s Somesh Saxena

Hello, and welcome to the latest edition of the digital CXO Leadership Inside Series. I'm your host, Mike Biard. Today we're with Soma Saxena, who is CEO for Panto Math, and we're talking about data observability and well, maybe the lack thereof.

So much welcome to the show. Thank you. Thanks for having me.

Everywhere you go, people are, uh, connecting the dots between the quality of the data and the experience they're able to provide and their customers and end users. And this is at the core of a lot of these digital business transformation initiatives. And yet we don't seem to have a good handle on our data.

We seems like we can barely manage it, nevermind monitor, and observe it. So what does it take to actually provide something that feels like true data observability? Yeah, very much Rick, Mike.

I mean, today, if you think of the data landscape and the data analytics world, it's pretty desperate. It's disconnected, right? There's so many different tools, technologies, you know, that move data, transform data store and house your data, visualize it.

And there is no horizontal representation of those cross-platform data pipelines that power data products and data reports, let alone, to your point, the ability to manage them correctly or monitor the data correctly. And so when you think of data observability, I think there's two critical components. Um, one of them of course is the ability to monitor things, but the other one that, that I think folks miss pretty often is just understanding what that end-to-end data pipeline looks like, right?

And that's where the company is named Panto Math. It means the one that knows it all. So what we do for data observability is we snap in across the customer's data stack for large enterprise and Fortune five hundreds.

Again, that's not just one or two or three hops, that's eight or 10 different technologies. We snap in and get an understanding of what those end-to-end data pipelines look like. That's in a fully automated fashion that we understand every interdependency and every relationship mapping from data producers to data consumers.

And once we have a holistic understanding of what those pipelines look like in real time, we then apply monitoring to it. And so now you're monitoring every granular asset in the data pipeline for missing data, stale data, uh, latency issues besides, uh, the most obvious things like data quality issues and job failures. But you're monitoring these things not in an isolated fashion.

You're monitoring that with a context of that horizontal end-to-end. So you understand as the data of reliability engineer, operations engineer, support team member what's broken, where it's broken, why it's broken, meaning root cause, and all the other impact points of running to remediate, to bring those pipelines up and running, refresh everything end to end, uh, with the context of, again, whose impact downstream and how do you make sure, um, they have light of statute of resolution so they're not making poor decisions with bad data. You can drive a trust in data and data transparency culture and avoid poor decision making with bad data.

I think most it people at least intuitively understand that there are issues with the data. It seems like the business people don't and are often surprised by the quality issues that emerge and sometimes just how done right, conflicting all the data actually is. Yeah, that's, that's very much the world I think most folks live in is because, I mean, you know, uh, data consumers, whether it's data analyst, whether it is a, um, you know, uh, traditional business user, uh, in, you know, finance, sales, marketing, any function, uh, and, and even executives for that matter, uh, their job is to just use the data and make data-driven decisions off of it in this context, at least in the data context, right?

They have obviously a lot more to do, but in the data context, they're just data consumers. To put it simply. They don't necessarily have an appreciation, uh, for the complexity behind the scenes, right?

They're thinking it's a report, maybe a tab lower power BI type for dashboard that is built for them. What they don't understand is that data is coming from so many different source systems, right? It could be a CRM, an HR system, a supply chain system, a finance ERP, and that data is coming in, um, on a daily or weekly basis, sometimes in real time.

Um, and then it goes through so many different hops. It goes through stitching, cleansing, transformation orchestration, uh, to ensure it gets to the data consumers at the right time with the right quality. Um, and again, there's so much complexity, so much technical debt that's built over the years with lack of data governance.

If, you know, the data space evolved, it's evolved so much in just last decade, decade and a half, uh, going from traditional ETL and data warehouses to now so much more complex, uh, uh, data ecosystems. And so, yeah, they don't necessarily have an understanding or appreciation for everything that goes on behind the scenes that the IT teams and data teams have to go through. Not sure if it's if if they need to or if there's job to, but it's on the IT teams and data teams to actually simplify that for them, is my opinion.

It's the other way around, right? And so, and so that's where I think Panama has, has, uh, shown tremendous value to their customers is, is as opposed to learning from the data, consumers reactively that there's a problem. 'cause they're caught off guard.

You know, sometimes they're preparing something, uh, with a spreadsheet they get in their email. Uh, sometimes they're looking at a data report in a meeting, um, getting called out in a meeting saying, Hey, something's wrong. Look at these numbers, right?

And you're talking to A CMO or CFO, and that's embarrassing. And then again, reactively, they're getting, getting to the IT teams like, Hey, what's going on? Why is this that hard?

Why is the data quality that tough to crack down and get right for the first time? Um, because again, they don't have that understanding or appreciation, right? So instead of flipping that its head and saying, you know, as opposed to being reactive, it can be proactive.

You can tell the data consumers in real time, Hey, don't make a decision after this report. Something's broken 20 half upstream. And then, uh, subsequently, hey, it's good now you're good to use it.

The health of that, uh, data report and everything leading up to it, uh, is, is shown to them in real time and they don't care about the rest of you know what's wrong and how to fix it. That's more the technical side of the world. And therefore those guys today, again, it's, it's pinpointing the root cause, uh, explaining to them the impact and giving them a guided path to resolution for that down the road.

The vision is a lot more than that. We can get to that hopefully in a few minutes, but, but even today, I think it's trying to merge those two worlds of data consumers and the data operators in a, in a very seamless way that doesn't exist today besides PAD to math. In my mind, this all comes back to a fundamental disconnect that's existed forever, but it people manage data and to them it's pretty much all the same.

And business people enter data and they theoretically are the ones that benefit from it, but they don't really think a whole lot about what that data means or its value or how it actually reflects something. And, um, this has existed as an issue for as long as I can remember. We finally kind of, kind of bring all this together in a way where everybody figures out that, you know, we need a more holistic approach here.

Yeah, absolutely. And that's where you'll see, you know, a chief data officer role becoming a lot more prominent in the last handful of years than it has been in the past. Um, because to your point, the IT teams, um, the technical teams don't necessarily have the domain expertise.

They don't really have an understanding of what the data really means, um, uh, and especially at the granular nuanced level. Um, whereas business users and data consumers do have an appreciation for, for what the data means and what value it can drive, but without necessarily at all times a holistic understanding of what that means for the company or the enterprise. Uh, right.

It's, it's very kind of siloed off to their vertical, right? So again, a sales or marketing or finance person is looking at their domain. A finance person can, can close the books with that data, but, but the, the larger impact of, of what that data could do and how you can get value out of it may not necessarily be realized in the, in the traditional context.

And that's where, um, a chief data officer role, uh, not only sitting sometimes within IT or within finance, but sometimes sitting at the larger leadership level reporting to A CEO is, is what's getting more common. Because data is a first class citizen today because, uh, of phrases you and I both heard is data is the new oil, right? So the value that we can get out of data, the potential at least that exists, um, is, is front and center of everyone's mind.

And now with ai, of course, it's, it's, it's even more prominent because, uh, until, until more recently it wasn't, it wasn't seen as that. Well, to your point, data is the new oil, but it feels like we don't have any refineries to actually make something with it. But, uh, um, but let's jump into AI here for a minute.

Um, is all this coming home to roost? 'cause every time I talk to somebody these days about their AI issues, the challenge is something is amiss with the data and it's not aligned, it's not prepped, it's not ready to be exposed to a bunch of algorithms. So is all this finally coming to a head?

It absolutely is. It, it, you're spot on there because you cannot have a good AI model. You cannot have good, uh, output, whether it is an AI model or a traditional machine learning model, or just historical trends that you wanna look at with no predictive component to it, it, it, it's really the end result's gotta be the same.

It's, uh, it's a, it's a common, uh, phrase in, in, uh, enterprises crap in crap out, right? Is, is that if you don't have good data, if you don't have good quality data, you're not gonna get good results. And so now with AI and the opportunity, uh, ahead with, uh, AI that most companies recognize data quality is front and center is top of people's mind because they realize that, that if they're gonna invest millions of dollars in ai, it's not really gonna move the needle for them without investing in data quality, in data reliability, and having observability and traceability across their entire data ecosystem.

In fact, when I look at the leaders in ai, the one thing they all seem to have in common is they had a fairly good structure for their data in place before the AI came along. And a lot of the companies you see in say, uh, pharmaceutical or finance, because there was regulations, they had a lot of their data in, in some format that was easier to expose to AI than a lot of other industries. So are we gonna see more industries being ahead of other industries just because of the fact that they were lucky enough to have their data in, in shape?

Uh, yeah, I think, I think I'm, uh, from the little bit exposure I have to that I'm, I'm, I'm noticing that too. Um, there is certain industries that are, that are, uh, you know, good early adopters because they aren't necessarily regulated the same way. And you're, to your point, in some cases because they were regulated, they weren't a good spot.

In some cases, the opposite. Um, if you're, if you're a healthcare provider, you know, it's, it's a lot tougher. You are under a lot of scrutiny, um, uh, with a, with AI regulations.

And so larger enterprises, fortune five hundreds have internal AI governance bodies and committees, um, that any investment, they're RAK in ai and especially bringing in external vendors, any kinda SaaS platform that, that moves the needle for them on the AI initiatives, um, is going through a microscope right now. Right? Um, and in some cases, honestly, it, it doesn't matter if your data was in a good place or not.

It's, it comes down to how much can you, can you, um, how much liberty, if you will do you have with what you do with your data, right? And so, uh, because you are answerable to regulators as, as you mentioned. And so I think certain industries like, like healthcare are definitely, um, at least from what I'm seeing, you know, a little bit kind of behind the curve because they have to think about some of those things versus maybe technology on the other side is a good early adopter, like they would be, uh, in, in most cases, right?

So at the end there, there's, there's some in between. Uh, it also comes down, I think, not just to, you know, how well were you positioned internally with your data management practices or you know, what regulations you have to abide by. I think it also comes down to the culture of the company.

That's, I think, a very critical piece because you could, you could not have any kind of handcuffs around you. You could have the data in good place, but a lot of cases it comes down to, you know, what the culture of the company is and where they choose to make their investments and, and, uh, and those, the, the, the, the, the decision making processes, um, are in a company compared to what their competitor, uh, may do. That timeless matters a lot.

Uh, that first movers advantage and, you know, you know, to wanna tap into the AI initiatives, uh, it matters a lot in, in, in what I'm seeing. You mentioned regulations or is that going to essentially wind up changing the culture in a lot of these organizations because spaces where you didn't see as much regulatory effort made are now gonna be exposed to more regulations? 'cause they're gonna be more AI driven and more broadly applicable.

And it's hard to think of a vertical industry that it won't be regulated. No, you're, you're, you're spot on there as well. I, I think, I mean, you know, in the, in the, in the, we're, we're, we're around the two year marks and since this whole thing kind of came about, right?

And, and, and now we are seeing a lot more regulation, we are seeing a lot more governance around this in some cases externally, in some cases internally, um, within, within organizations. And it's, it's, I think there is gonna be guardrails, there is gonna be a bit of a, slowing down is probably the wrong phrase, but, uh, but just, you know, a lot more eyes on how to do this sustainably, how to do this more diligently. Uh, and, and that's I think the world we're living in today is, is everyone sees the opportunity with ai, but how do you go about it in a way that is, that is, uh, a lot more, um, diligent versus, you know, maybe a year, year and a half ago it was like, let's just do a quick POC, let's quickly go and see what we can accomplish with this, right?

But, but now it is, if we have to operationalize an AI model, if we have to roll something out across the enterprise, if it's gonna impact the healthcare example, you know, our entire, you know, customer database and, and you know, we're plugging all of that data into, into a model, um, yeah, we're gonna, we're gonna think about that before we move forward with that, right? And so that's, that's something the phase we're in. It, it, it does, I think get on guardrails in the next year or two after we've gone through a couple of iterations and folks just kind of learn this, this new world they're living in and, and what, what's right, what's wrong?

What are the do's and don'ts? Um, so I think it will kind of normalize, if you will, uh, and get more repeatable. But, but that, that's gonna take some time.

One of the dirty little secrets about it in general is that, you know, most business users, when they get a report from the IT department, they view it with suspicion. And their immediate sense is, you know, well, I know where the data came from so I know how flaky the report might be. And so they not inclined to make, uh, you know, a so-called data-driven decision.

And sometimes when it conflicts with their instincts, they, you know, will lean on their instincts. But that can be equally problematic. 'cause you know, that gut feeling might just be what you had for lunch.

Um, so my question to you is, as all this evolves, you know, are we gonna get to the point soon maybe where people do have more confidence in the data? Absolutely. Uh, what you're describing is very much, um, the challenge.

Almost every organization I have had exposure to through my journey in, in IT and data faces, um, there is no getting around that, um, without an investment in data reliability, without an investment in ensuring the data's accurate and healthy. Um, part of that is absolutely the technology investment where a tool like padma can can play a big part. Uh, and we have seen that, you know, play out successfully for, for several customers of ours who, you know, were living in, in the world where they weren't trusting the data business users weren't looking at the reports, even though they had built thousands and thousands of dashboards, people going off their gut field decisions and now trust their data a lot more.

Um, but, uh, but then the other piece outside of technology, uh, and anything a Panama could do comes down to the culture of the company, right? And that culture, uh, I think, um, it has to be fostered internally because, because, you know, we, we face this at GE as well, uh, where, um, general Electric, when, when I was the head of data and analytics, they, and even my, my journey leading up to that role, you know, the company went from, um, not being, it was always data driven as a company. General Electric, of course, with the Six Sigma Backward is always a data-driven company, but not necessarily always a digital company.

And so I've seen that transformation firsthand from, uh, general Electric going from, you know, we have a few reports, um, there's things in Excel sheets and people are kind of in their own silo off world, maybe going directly into an RP system, getting extracts and, you know, number crunching themselves with their own formulas and trying understand what they do there. Um, two, let's now build reports across the enterprise in a more repeatable fashion and a more sustainable fashion, um, in the data lake and data warehouse ecosystem. Uh, so it's not person X who did X calculations versus person Y, and they're coming to different answers.

No, they're looking at the same report. They have consistency and clarity on what the numbers mean and how they got those numbers. And based on that, they can make a more informed decision for the company.

So some of that I think is a culture piece regardless of, um, any technology playing a part. And I've seen that evolve. I've seen General Electric be extremely successful with that.

Um, uh, and then comes I think the technology piece, which is you can, you know, build complex data pipelines, build amazing dashboards, drive adoption, drive a trust in data culture. Um, but that will be, be, uh, dampened by not having reliable and trustworthy data without a data observability platform like Panto Math, Right, folks? Well, you heard in ears, Hey, the truth is out there.

It's in the data somewhere. The problem is, is there's too many forms of the truth right now. So we need to sort it out.

Hey, Esh, thanks for being on the chat. Thank you. And thank you all for watching the latest episode of the digital CXO Leadership Insight series.

com. Until then, we'll see you next time.