In this episode of the AI Data Fabric Show, Prat Moghe speaks with Neil Bhandar, Chief Data Officer at Generac, about his career path across multiple industries and functions. Neil has served as CDO in diverse contexts—from launching Tide Coldwater at Procter & Gamble to leading HR analytics at JP Morgan. He describes his approach as treating data as a universal language where the “accent” changes depending on the audience, whether that’s supply chain, marketing, finance, or human resources.
The conversation covers both opportunities and challenges in AI, including the risks of multiple organizations using identical pre-trained models and the analytics literacy gap among executives. Neil discusses why the cost of answering each new question in analytics remains high, how he approaches understanding business problems in different functions, and why he considers himself a “business person who sits in the data world” rather than a data person trying to understand business. The episode offers perspectives for CDOs and anyone working to connect data initiatives with business outcomes.
Watch the full episode below or listen on your favorite podcast platform (Spotify, Apple).
Full Transcript of the Episode
The AI Data Fabric Show
Guest: Neil Bhandar, CDO at Generac
Host: Prat Moghe, CEO of Promethium
Prat Moghe: Hi folks, this is Prat Moghe with the AI Data Fabric Show. I’m super excited today—with my cup of chai—to welcome a really accomplished guest, Neil Bhandar, who’s the Chief Data Officer at Generac. Neil has been a CDO for many companies and has had a really interesting and remarkable career doing this across different industries and business functions. I haven’t seen somebody with that kind of versatility, so I’m really excited to welcome Neil to the show. Welcome, Neil.
Neil Bhandar: Thanks so much, Prat. Thanks for having me. Delighted to be here.
Prat Moghe: Awesome. So Neil, one of the things I mentioned is you’ve done companies both small and large. You’ve worked at Procter & Gamble in strategy, planning, and marketing—that’s serious stuff. You’ve launched your own startup, which is also super impressive. Then you moved to financial services at JP Morgan, where you took on the Chief Data Officer role but also took on a function like HR, which I find super interesting. And then you’ve done the horizontal thing—you’re now at Generac, which is a very different space and sector.
It seems like math is a theme that goes through everything you do. You teach statistics at a grad school level on the side in your copious free time. So my first question is: what got you started? How did you get into data and analytics? Was there a specific point where you said, “Look, I want to do this”?
Neil Bhandar: Yeah. I was always a fan of math. I grew up in India, and the culture was you either became a doctor, engineer, or lawyer. So I had similar inspiration from within the family and was drawn into engineering. My dad was an engineer, so that’s basically how I ended up becoming an engineer. The math and the numbers were always exciting and interesting.
When I finished my mechanical engineering degree, my interest was in math, but I wanted to stay close to engineering. So I decided production planning, sequencing, scheduling—the heuristics behind it—fascinated me. That’s how I picked operations research. The OR draw for me was looking at logistics networks, various ways we could do optimization, and where problems are more complex, what other alternatives existed.
What ended up happening over time is I started seeing a lot of similarities in problems. I often give my kids this example: early on I was in supply chain planning, and my focus was inventory optimization—multi-echelon inventory optimization. If you think about a classic inventory problem, if you plot this on an XY axis, it’s a very typical sawtooth. You get a replenishment, your inventory goes up, over time it depletes. I started to notice that that triangle repeats itself in multiple different areas.
I’ll give you an example: when I was in financial services, one of my very first projects in the credit risk space was in the credit card business. Believe it or not, that same triangle shows up all over again, but now in a very different industry in a whole different context. When you take a credit card, you start with zero balance, over time you increase it, and at the end of the month you pay. It’s the same sort of thing—just the peak happens at the end versus at the beginning.
Now, the geometry of the triangle becomes very interesting. Triangles are very difficult shapes to focus on, so what you end up doing is taking the area under that triangle and converting it into a rectangle by taking half the base times the height—which is nothing but half of that triangle converted into a rectangle.
Over time, if you’re solving an inventory problem, you’re trying to figure out what your carrying cost is going to be and what your cost of replenishing and placing an order will be. In the financial services world, you’re trying to figure out what your loss rate is going to be for those people who are going to walk away with your money—because their triangles don’t look even, right? They carry a balance. Or you have a cost of capital because you’re lending money out and there’s cash coming in. At some point in time, that loss is going to have to catch up, so you have to price that in.
Long story short, what interests me about math and numbers is being able to visualize these as geometric problems and then manipulate them in my head. I love doing this. Often I wake up with some of these visualizations in my head early in the morning.
Prat Moghe: That’s like a deep research agent now in the AI world. That’s great.
When you started, one of the things I found fascinating was Procter & Gamble is known for its depth and rigor when it comes to strategy, planning, marketing, and analytics. You were part of the group that launched Tide Coldwater, right? Can you share the concept, the result, the impact? Why did it succeed? Anything you can share?
Neil Bhandar: It’s funny—the typical perception of what a brand marketer in any organization, not just Procter & Gamble, is that they’re more touchy-feely, more emotional, more communication-oriented. The reality is actually a brand marketer is a strategist. They’re all about data. They’re all about trying to figure out how to make sense of it in the real world. As analytical as they are, they’re also trying to figure out how to make money out of that data, how to generate value.
Back in 2004, we were sitting on an innovation that had been around for quite a long time. The idea is when you put warm water in your detergent, the detergent acts more effectively toward cleaning. At Procter & Gamble, the innovation was they had an enzyme-induced detergent on the base side that didn’t need activation from hot water or warm water. The catalyst that warm water provided didn’t need to happen—it activated on its own the moment you put water into it.
We were in a cycle back then—this was the second Bush administration—and the cost of a barrel of oil went from $45 to about $105. People were feeling the pinch; inflation was high. So we realized this was an opportunity to take this innovation where you could use cold water in your detergent and save the consumer some money. We’d done some research—I think back then it was something in the order of about $63 per year. Not a huge number, but as a fractional amount, it’s significant.
The other thing was when you use cold water—and you’ll notice this with some of your finer fabrics and sweaters—they highly recommend you use cold water so the clothing doesn’t shrink. And the last piece is when you use warm water, there’s a greater tendency that colors will bleed.
So we had a very clear set of value propositions that we could present. We took all that information from our communication perspective and said, “Look, these are the kinds of things we want to emphasize and highlight.”
Now, I will tell you there is a downside to this too, by the way. The equipment makers—the Maytags, the Whirlpools, the Samsungs of the world—hated it. I’ll tell you the unintended consequence in a second.
What ended up happening for us was we took this three-pronged value proposition for the consumer from a communication perspective. Tide as a product is a loss leader, so what that means from a CPG perspective is if you have a category captain, retailers will sell the product below a certain price where they lose money on it. The intent is because that product is on promotion, it brings crowds into the store, and ticket size and basket size are high. They make their money on volume, so they’re willing to forgo the profit on that product to get the rest of the purchases that happen.
We were aware of that fact, and we were also aware that if we continued doing that, there would come a point where the next product that’s better than Tide would replace Tide. So we wanted to constantly seek out opportunities by which we could build some margin for the retailer and pass that margin on.
Back then, one of the strategies we pursued was “down-counting.” Down-counting means if you take a base-size product, it’s a concentrated product. You’re not required to fill the cap to its brim—you actually only need to go to the lowest graduation and you’re still fine. But people still tend to overdose.
What we did was we down-counted it in terms of concentration, kept the pricing the same, and as a result, the margin that we built on it, we passed to the retailer. So it became a profitable product for the retailer. That was the second prong.
The third prong was we basically said, “Let’s focus on those retailers.” When we launched Tide Coldwater, it was a very small budget. Typical budgets at Procter & Gamble are in the $100-plus million range, but the smaller budget meant we had to be very conscious about where we focused from a shopper activation perspective.
At Procter & Gamble, there was this philosophy back then: “first moment of truth” and “second moment of truth.” First moment of truth is when you’re standing at the shelf, looking at that expanse of all the different products available, trying to make a decision. Often that decision is made by your brain—it’s not quite an impulse decision, but it’s typically what, in Daniel Kahneman’s words, is when your System 1 and System 2 are playing with each other. You’re making a decision not with your heart but with your head—it’s a cerebral process.
We wanted to make sure that when that happens at that point of retail, we wanted to give them a compelling proposition. So we decided on focusing only on those retailers and channels where we had high close rates in the category, we had “golden households” that tended to shop in that category more often—and a golden household for that business was people who were going to do more than one load or maybe even multiple loads a day, especially people with young kids—and those retailers where we had resonance with our prime prospect for the brand.
That gave us that three-pronged firing approach, and as a result, in a matter of less than four weeks, we were able to gain 4% market share. That’s very significant for a $7 billion brand.
Prat Moghe: That is awesome. I mean, that’s just like the power of analytics. But thinking back, you basically got a very clear ICP—ideal customer profile—driven by intent. It goes to show that if you combine analytics with an understanding of the business… It sounds like you were super comfortable—as somebody who’s trained as an engineer, techie, statistics, data science—you were very comfortable actually going and hanging out with the business people.
This is one of the challenges I’ve noticed with many CDOs. If they come from a technical background, they have a hard time figuring out how to team up with the business. Anything you can share there that helps all the CDOs and CDO aspirants out there?
Neil Bhandar: Yeah, and that is typically the way I describe myself: I am a business guy that sits in the data world, not the other way around. Believe it or not, this is my very first role at Generac—throughout my entire 20-plus year career—where I am positioned in IT. I’ve never been in IT.
This is generally how I describe it: data is data is data is data. It’s like English—no matter where you come from, it’s still the same English, but the accent that an Australian will have will be different from a British person, from an American. Even within the United States, if you’re in the northern part versus the southern part of the US, data is the same way. The accent changes based on where you are—whether you’re a data person in the business and that business is HR, credit risk, supply chain, or marketing. But you’re still dealing with that same data.
I think for all aspirants in the data world, you need to figure out how to modulate that accent. When you’re having a conversation with a supply chain person, you need to speak supply chain. You don’t need to speak AWS, Azure, mixed integer linear programs, heuristics, or genetic algorithms. You need to speak that language with that particular audience. You need to be able to—not to make it sound like a chameleon—but really adapt into that business. What are their pain points? What are they looking for? What will get them to, one, trust you, and two, start realizing value quickly?
Prat Moghe: So let me ask you—this is a great point. You moved from marketing—I’m using marketing as a loose term—but you moved from marketing into risk, and then HR at JP Morgan, for example. When you went to HR, that’s very interesting. How do you deconstruct or understand what the important problems are that they’re trying to solve? Ultimately, you have to bring this back to—like you said—triangles, optimizing these things. What was your observation when you landed in HR, the way they were doing it, what was important, and how does analytics ultimately impact HR?
Neil Bhandar: This is typically independent of HR or any other function—how I approach it. Before I go into any function, I create my own preconceived biases and notions of what I think are important. In the first few weeks when I sit in those meetings with the audience, I’m trying to validate some of those assumptions. If they validate, great. If they don’t, there’s a new learning for me.
When I first went to HR, I thought all they did was hire, fire, and pay. That was my understanding—I had very limited understanding of what HR did. Believe it or not, when I first went to financial services, I had a very limited understanding of what financial services was.
Long story short: as you get to a certain milestone within the first two, three, four weeks and you’re sitting with the executive leadership, it became obvious HR was most interested—as a cost center—in what they’re paying for, what’s incurring cost for them. Whether it was attrition, whether it was the cost of hiring people, whether it was engagement that wasn’t necessarily translating into active value.
We started looking at: there are people in this business, but they’re not talking to this business. So we started looking at network graphs of who’s talking to whom. We led some organizational network analysis. Then, what will give them the incentive to talk to one another? Can you embed champions from one part of the business to the other?
As you start to pick up on some of these signals, then you start to think, “Okay, what is the mathematical way I can handle this? What will give them the solution they’re looking for?” They don’t need to know it’s ONA—organizational network analysis. They just need to know, “I need to get this part of the business to talk to this part of the business. I don’t know what is going on and why they’re not doing this.”
Prat Moghe: Your job is to model that process.
Neil Bhandar: Correct. The other thing is HR, believe it or not, is an extremely data-oriented function. The fundamental challenge for them is they don’t trust data. So they’re all about, “Yes, give me the insight, give me the analytics,” but at the end of the day, “I want line-by-line information that I want to keep in my back pocket. If somebody asks me how did you arrive at it, I can rattle off names.”
So largely they are still in a world where they’ve been burned maybe a few times, and as a result, there’s a trust gap that exists. The good part was, given they were data-oriented and were already using stuff, it became easy to address those trust issues in order for them to start embracing the more advanced analytics.
We started getting into this—there was this whole exercise that my team helped with called employee opinion surveys once a year. Every quarter we did what are called pulse surveys, which are shorter versions of the employee opinion survey. Then the question came up: “There’s all these anecdotes that people have—happy employees make happy customers, right?” So we started to get into the idea of validating, “Is that true?” Because these are like mini-bombs that would get dropped every so often.
Prat Moghe: It’s common wisdom.
Neil Bhandar: Correct. What we found was that there was no data that actually validated that. Look, one thing I strongly believe is if you torture data enough, you can get it to say whatever you want it to say. You can take a scatter plot that’s all over the place and take a segment and say, “Look, here high engagement, high satisfaction works,” but then you’re ignoring the rest of the dataset that exists.
So there were some of those sorts of exercises that we went through where we said, “Look at your data not just from the perspective of new discoveries, but validation of existing biases.” If this bias exists, is it really true across the universe, or is there a pocket where it’s true? That became an exercise where we started going after small pieces here, small pieces there, and chipping away at it in a way where we started building more trust with the capability that we were bringing to the table.
The same happened with questions like when people said, “I’m looking for a job and I don’t see myself with the company in 12 months’ time.” It was a typical question: “How often do you look for a job outside? Do you see yourself with the company?” We started analyzing that information to see if that was true, and largely people were very honest in answering that question.
So then we started saying, “Okay, what happens to those people that answered yes but still stayed with the company?” We quickly found out that when you look at their longitudinal history with the company, they changed managers, or they changed their teams, or they changed their function. Those sorts of questions were the ones that were part of conversations that were happening. There was that undercurrent of conversations which were drawing out for us the necessary threads to pull on.
Prat Moghe: That’s very interesting, and you can do something about that. If somebody says that, then it basically says maybe the environment’s got to change to keep them. You’ve got to write a book on freakonomics or something.
Neil Bhandar: It’s funny you say this, Prat, because I’ve been working on a book for a little while now called The Cost of Curiosity. The whole concept there is I believe the world of analytics needs to evolve into the equivalent of an Alexa device. There’s a very small fixed cost—in my case, I bought the device for like $29.99—and then on an ongoing basis, the variable cost is just internet and power. That’s it, nothing more.
We need to get to that stage with analytics and data. Today, the marginal cost of answering a question is huge. I ask a question today, and then there’s a follow-up question—it either costs the same or more.
Prat Moghe: And sometimes it takes so much time that the question has often lost relevance.
Neil Bhandar: That’s right. There’s no reuse. There is no—like you said—basically the cost of…
Prat Moghe: And we’re not even talking about the real cost of AI.
Neil Bhandar: Exactly. So there is definitely an opportunity to reduce that cost of curiosity. That’s what my book is about.
Prat Moghe: Awesome. Add me to your early review list. Happy to help you in any way.
Now, with AI, Neil—you started with statistics, machine learning, you’ve gone through all the classic algorithms, optimization, linear, nonlinear, all that, graph theory. When you look at AI and you look at both the challenges and the opportunities, what comes to mind? Give me some tough observations.
Neil Bhandar: Yeah, I think this is both interesting, scary, and fascinating all at the same time. One of the biggest challenges the way I see it facing most of us is we do not have a common baseline understanding of what is AI. When people talk about the risk of AI, they’re often fixated on the kinds of things they hear in media—like hallucination, for example. Those things are not really a problem that I have to face because we’ve got architectures by which we can control it.
What I am most afraid about is the sea of sameness. You use OpenAI GPT-5.0, I’m using GPT-5.0—it’s the same exact pre-trained model. The only difference is you now have some sort of vector database, some ontology that is controlling for hallucination. But the model that is helping arrive at a decision is the same.
There isn’t the depth of understanding of some of those concepts in the business. On a separate front, one of the challenges that we have is we are definitely in a hype cycle. Trust me, I am a big believer in AI and I still think there is a ton of value that it can create, but the hype cycle has a lot of blinders on people when it comes to decision-making on what needs to happen.
You’ve got about three-quarters or more of CEOs that graduated before the internet came to be. You’ve got 90% of CIOs who have been in the role before generative AI came to be. So you have a huge literacy gap that exists. As a result of it, when somebody says “we need more AI,” these are the kinds of blind spots that they are not quite able to navigate easily. That is a huge risk to me.
Analytic literacy is one thing that if we don’t bridge quick enough—and when I say analytic literacy, I don’t mean they need to understand what a layer of perceptron means, what back propagation means, what epochs of training are, what bias is in the back propagation sense. I mean those kinds of things—it’s not the technology, but it’s how it’s being used. And that sea of sameness, this piece that I’m talking about—that is scary.
Prat Moghe: 100%. I think partly it’s like, back to your point of explainability—why is something…
Neil Bhandar: Very much so.
Prat Moghe: It’s kind of like I give you a black box, I give you inputs and there are outputs, and if you’re not able to explain…
Neil Bhandar: Very much so how they’re tied. Even if that model is… I think your point is, firstly, there’s this averageness, and then the question is how do you stand out, stand away from that? What’s unique to you? How do you bring that?
I have one additional example I want to share with you, Prat, because this one is quite interesting. In the banking world, obviously there is this constant endeavor to make sure that your next model performs better than your previous model. There’s this constant champion-challenger dynamic where teams are working toward beating models when it comes to portfolio performance or creating synthetic portfolios and so on and so forth.
Often what people will end up doing is they’ll do second-order, third-order models. So input from one model goes into a second model, that model goes into a third model. There’s a huge risk of overfitting, in a more machine learning sense.
Prat Moghe: Correct.
Neil Bhandar: We have a model today where we are using a pre-trained model to generate a machine learning model that is then being deployed.
Prat Moghe: Agentic.
Neil Bhandar: Yeah, you’re just going to do that at scale. And can you imagine now—this is where the literacy piece comes in—resonance.
Prat Moghe: That’s right. Unintended consequences.
Neil Bhandar: So I want to go back to the unintended consequence of the coldwater piece: bacterial growth inside your dishwasher, your washing machine.
Prat Moghe: Oh, which hot water would have taken care of?
Neil Bhandar: Correct.
Prat Moghe: Very cool. But was that known?
Neil Bhandar: Not to us. But when the folks at Maytag, at Samsung started complaining, it became obvious that this was the risk. So now you can sell them another product, which was—
Prat Moghe: Well, obviously.
Neil Bhandar: But the point basically is some of these unintended consequences of these sorts of things. Now think about this when you’re deploying these models to make very large autonomous or semi-autonomous decisions. You do not know what that resonance is going to look like when multiple models that are independently running—without the knowledge of the other existing, coming from the same pre-trained model—is going to result in resonance that could be devastating.
Prat Moghe: That’s right. You’re running very fast in the wrong direction.
Neil Bhandar: Exactly. And you don’t know. That’s the scary part.
Prat Moghe: What is exciting to you?
Neil Bhandar: Oh, the technology is fantastic. For the longest time, there was this huge scare that there isn’t adequate talent. Now we’re facing the exact opposite. We’ve gone to 15,000, 20,000 layoffs of knowledge workers, white-collar employees, because you’re now suddenly realizing that they can be replaced with an agent.
It’s not the fact that we’ve addressed that lack of resources, lack of talent gap—but the possibility of what this technology can do is absolutely exciting. I often say this to folks: I feel like we are at that stage where we are truly limited by imagination and physics. I felt like you couldn’t have said that with enough confidence up until now.
Prat Moghe: That’s right.
Neil Bhandar: Any parting advice to CDOs, people in the data analytics space, particularly the ones that are straddling data, IT, and business?
The few things that come to mind: First is, if anyone can take anything away from my personal career, I’ll say agnostic to industry, agnostic to function, agnostic to geography—data is universal. You are certainly limited by your curiosity, your interest in learning those things. Just because you’re in data doesn’t necessarily mean you fit everything and anything. Your fit depends on your curiosity.
Prat Moghe: Absolutely. That was awesome. I really appreciate your time, and I’m looking forward to reading your book when it comes out, Neil.
Neil Bhandar: For sure. You’ll be the first on the list. Thanks so much. Have a great weekend.
Prat Moghe: Thanks. Bye-bye.
SHARE THIS:
