Creating an Intelligent Future by Blending Tech Innovation with Real Business Impact

UWEBC Podcast Introduction

Customer experience, technology, people, and supply chains are shaping how organizations operate today, but keeping up isn’t easy. At the UWEBC, we connect leaders navigating transformational change every day. Each week, we bring you the journeys and perspectives of Wisconsin leaders shaping the future of business.

From the University of Wisconsin–Madison, this is the UWEBC podcast.

Introduction and Setting the Scene

Doug Barton:
Hello, everyone. It’s Doug Barton from the UWEBC, and this is another episode of, in this case, Catalysts at the Capitol Getting…

What are we getting?

Karthik Josyula:
Chai.

Doug Barton:
Latte.

I’m here with Karthik Josyula.

Karthik Josyula:
Yeah.

Doug Barton:
Did I get that right?

Karthik Josyula:
Yeah. Very first time.

Doug Barton:
I did. I meant to ask you that before. That’s the way it goes, though.

This is a special episode. I’m actually here with a notebook. We’re outside the car on a beautiful Saturday. The Dane County Farmers’ Market is happening all around us, just outside our State Capitol.

It’s extra special for that reason. We have our first and most technical expert with us today, Karthik, who is going to share some of his background and experiences as they relate to creating an intelligent future.

I’m super excited to have him on board, so let’s get right to it.

Meet Karthik: Background and Education

Karthik, why don’t you give us a little introduction and bio, including where you studied and what you’ve been up to?

Karthik Josyula:
Absolutely, Doug.

I did my schooling in India, mainly focusing on math and physics. I did my undergraduate degree in electronics and communications, mostly focusing on satellite communications, semiconductor devices, and so on.

Professional Journey: From India to the U.S.

Then I started working for banking and financial institutions in India.

My launchpad was mostly focused on the .NET Framework, building applications with the .NET Framework and ADO.NET, which was the data layer, integrating databases and the business layer within the operating frameworks.

After that, I became closer to the data, where I identified how important it is to build an accurate data layer because that drives insights.

The old-school reporting still had value. I believe we used to have SSRS reports, and most of the reporting was done in Excel sheets, but a metric is still a metric.

That’s where I started my journey toward data. I’ve been a data architect, and then I came to the U.S. in 2014. I worked for BMO Harris and Northwestern Mutual in roles including architect, chief architect, and strategy consultant.

Then I wanted to expand my technology knowledge, so I started planning to complete my master’s degree in data science from Northwestern.

It was a very interesting learning experience that changed my perspective toward life because I was working a full-time job, I was a husband, I was a dad, and I was also doing my coursework with weekly assignments.

That gave me a structured understanding of how to approach life and balance priorities well. That was a fabulous lesson I got from data science.

Doug Barton:
Totally.

We didn’t talk about this, but that was kind of my experience, too. I’m grateful to my wife for supporting our family as I went back to school. I did an executive master’s program at Kellogg, at Northwestern, which is a terrific learning community.

I’ve been fortunate to experience Wisconsin, which I adore, of course, and where I spend my time working now. But that learning community in Evanston is something extra special.

That notion of work-life balance, or work-life integration, is really essential today.

Karthik Josyula:
That’s exactly right.

Doug Barton:
Good. Let’s talk a little bit about your work experience.

Where are you now?

Karthik Josyula:
Right now, I’m with Kohler. I lead the data and AI platforms.

I work on the data strategy, redefining the architecture and making sure our data is sitting in one place, accurate, reliable, and driving all the important key metrics for the business.

I also lead the AI platform practice, where we create accelerators, do rapid prototyping, manage the models, and maintain the models.

Cost is a very important metric that we cannot turn a blind eye toward. We need to operate well within the allocated cost, so we also do cost-optimization initiatives.

Doug Barton:
I think that comes with the maturity of any technology. We don’t just care about performance, but cost and performance in a very holistic sense.

Before you move on too quickly, Paul Ryan, chief digital officer at Kohler Co., was on stage with us at our Wisconsin Digital Symposium. He reminded the audience that it’s about a 150-year-old company.

I have to imagine there are many varied data sources there as you try to create a platform over it.

Challenges and Innovations in Data Management

Maybe share a little bit about how you think about getting started, maturing, and innovating, perhaps all three in different spaces.

Karthik Josyula:
That’s a journey.

First, any organization needs to evaluate where it is right now and what pain areas it wants to solve.

Do they know where the data is? Do they know where the data is, but it isn’t accurate? Do they know where the data is and know it’s accurate, but it isn’t driving the key insights?

An initial evaluation is much needed. That’s exactly what we did at Kohler.

We knew the data was in place, but our data-availability metrics were not that great. Month-end reports were always a challenge to deliver on time.

We evaluated where we were, what bottlenecks were happening, and then completed a technology evaluation. After that, we started moving strategically into a modern architecture.

Now the modern architecture is in place, and we are completely migrating from our old environment to the new environment.

It’s still in progress, but now that we have enabled a catalog, that will help my data-governance friends build certified datasets.

Now we know where our data is. We clearly know what key metrics the data is driving and what the accuracy of the data is.

We have addressed all three. Now innovation will be accelerated because the data is known, the accuracy is well taken care of, and we know what key metrics are being driven by these particular data points.

Doug Barton:
Let’s follow that thread on innovation.

My research, which of course means my AI-assisted research about your prior work, uncovered a phrase that I think said, “from model to money.”

ChatGPT on your phone and deep research are amazing tools. I spent some compute to do this.

From Model to Money: Translating AI into Business Value

How do you translate the power to predict into business value? I think that can be misunderstood. Share a little bit about that.

Karthik Josyula:
Absolutely.

The term or phrase you used, “model to money,” has a predecessor to the model, which is the business problem.

Doug Barton:
Indeed.

Karthik Josyula:
Business problem to model to money.

First, we need to identify what problem we want to solve and what ROI we would get if that particular business problem were solved.

Once the business problems are put on paper and prioritized, we want to go after maybe the top three.

Why the top three? The top one might be challenging with respect to technology. It might be a very difficult problem to implement.

We could keep innovating and doing proofs of value on that first problem. Instead, maybe take the top three or top five problems, distribute them to the teams, conduct proofs of value, and bring technology in to solve the problems.

Not everything should be solved with generative AI. A simple statistical algorithm can solve a couple of challenging problems.

Doug Barton:
That has been an interesting challenge.

Generative AI captured people’s imagination and attention, but we can’t forget about traditional predictive AI and classification.

There’s so much to do and unlock. I think we as technologists are probably grateful for the attention, but we also should be mindful of matching the right problem with the right solution.

Karthik Josyula:
Exactly. Bringing the right solution to solve the right problem is very important.

Now that we have proved the value with our models, we put them into operations.

We deploy them, monitor the models, put a cost equation around them, and put privacy, security, and governance guardrails around them.

Then we can start realizing the value we can extract from these models that are trying to solve the problem. That’s when we get the ROI.

Doug Barton:
When we sat down a little earlier, I flashed this book up.

Karthik Josyula:
I’m excited to read this one.

Doug Barton:
Good for you. The print is big, I have to say, but it does have 388 pages.

The graph in particular that I want to draw your attention to walks us through this three-part story about how the value-creation factors come together to create value.

There is always this notion of feasibility, including what is technologically feasible and approachable.

Karthik Josyula:
For the audience, I’ll try to explain it.

On the x-axis, we have technology feasibility, or feasibility risk. On the y-axis, we have the business opportunities. There is also a sloped line, which is adoption risk and should be treated like the third dimension.

We have three dimensions: feasibility on the technology side, business opportunity, and adoption.

Doug Barton:
It’s only when something is technologically feasible, delivers business value, and gets adopted and used that we can ultimately have the impact and reach that potential.

To your earlier point, adoption is the ignored metric.

Karthik Josyula:
Yes.

I think it should start with the business problem and the business opportunity. Evaluating that is very important.

If a prototype or solution is in place, do we have protocols or policies that will take care of the people and the change management?

I think it’s a parallel initiative that should begin alongside the technology.

It’s not that we solve it with technology and then figure out what the people and process changes look like. The parallel work should start right away.

They should meet at one point where the business value is proven through technology, and we immediately have all the policies in place and understand what it takes for the change to happen.

Doug Barton:
That will speed adoption.

You mentioned the notion of guardrails a little earlier to keep us within responsible innovation and responsible governance.

There are a couple of topics I jotted down here, and they can be thought of as a two-part story.

Developing Talent and Future Trends

Talk a little bit about developing new talent and what talents you think are important.

Is it both technical proficiency and a sense of curiosity about business outcomes? It certainly seems like it should be.

There are also some next big things coming. We were talking a little bit about the Data + AI Summit that Databricks runs. Maybe you can talk about some of the things they’re doing to enable that.

Let’s start with talent.

Karthik Josyula:
I’ll begin with talent, including young talent and how to grow talent.

The first key thing I look for is enthusiasm. How excited is an individual to solve a problem?

Again, we are not talking about generative AI or the latest and greatest agentic AI capabilities. It could be as simple as data exploration, finding outliers in the data, or finding anomalies.

How excited is an individual to find the outliers or anomalies in the data, do anomaly detection, and tell a story out of it?

That curiosity is very much needed to ignite talent.

The second one is ownership. Is an individual ready to take ownership?

No one likes half-baked solutions or half-finished work.

I have a funny story. We had a water leak in our living room. We called a plumber, and he came in, made a cut, fixed everything, and left it open.

I said, “Can you close it?”

He said, “That’s not my job.”

No one likes half-baked solutions or half the job.

When I draw a parallel to our work, once a task is taken on, I want the individual to deliver it through the end. Having that sense of ownership is very important.

Doug Barton:
I think there’s something very human about wanting our technical mastery of a topic to have an impact on people.

The unlock is showing that your expertise, which you’ve developed and mastered through your experiences, actually has a payoff for other people and their lived experience.

That’s very cool.

Let’s talk about enabling those individuals. You talked a little bit about encouraging the people side and the mindset, but how do we uplift students or young talent so they can work closely with the technology?

Karthik Josyula:
That leads into the second topic, the Databricks Summit. There were some very exciting announcements there.

Databricks announced a free account, a lifetime free account.

I think that’s a game-changing capability, especially for young talent and students in school who want to work on data.

Doug Barton:
Take a moment and describe Databricks.

IBM, where I worked for many years, and Databricks were the top two contributors to the Apache Spark project.

But unpack Databricks. It’s not the Databricks we knew 10 years ago.

What does a slice of Databricks mean to young talent? What would they have access to there? Maybe share a few of the features.

Karthik Josyula:
First, I’ll explain what Databricks is and what they do.

I’m not from Databricks, but with the knowledge I have, I’ll try to share what Databricks is in layman’s terms.

I won’t go deeply into the technology, and then we’ll return to your question about what the audience should know about Databricks.

First, Databricks manages clusters. They provide infrastructure for data processing and ETL steps.

Doug Barton:
Sometimes we think of clusters as compute. Isn’t that a term that’s used synonymously?

Karthik Josyula:
Yes, compute clusters.

More clusters mean more compute. Every cluster has a segment. For example, a cluster could have four gigabytes, or on the other side, a cluster could have 16 gigabytes.

It’s compute. Compute is what the power costs.

In layman’s terms, people use the terms interchangeably. Saying, “I want a high cluster,” means they want high compute.

They have minimum working cores and maximum working cores, which are adjustable.

Going back to that question, they manage clusters.

Optimizing Cluster Compute

They used to manage clusters. Then they identified the challenges businesses and organizations were facing when the data movement happened.

That’s when they brought in optimized approaches so the cluster compute could be utilized to the maximum.

That’s where Spark SQL and PySpark come in. These languages enable Spark compute to operate at maximum efficiency.

I can write SQL code or Python, and it still works in Databricks. But without the right tools, it may not leverage the cluster and Spark capabilities for optimized computation.

Doug Barton:
That goes back to something you started with very early in our conversation.

If we’re going to be responsible for innovations that matter and have an impact on the world, we have to think about cost, too.

By optimizing this stack, they’ve actually been able to produce a free slice of Databricks to allow people to experiment more.

These things are interchangeable in some respects. By pressing costs down, they can maximize access, and we’ll all benefit from that.

Karthik Josyula:
Imagine that with $10, I’m able to buy 100 gigabytes of compute.

Tomorrow, with optimization, the same $10 could buy 1,000 gigabytes, or one terabyte of compute.

That’s the balance between the dollars and the memory.

The Role of Apps in Business Solutions

Doug Barton:
Do you want to take us back to one of the other announcements, which was apps?

Bringing apps into the picture is important. Now that compute is available at low or zero cost, the notion of prototyping impactful applications could help with translation.

Karthik Josyula:
There might be a dotted-line connection between the compute cost and Databricks Apps.

The main interesting thing is the proof-of-value journey.

You previously made the statement “model to money,” and I added “business problem to model to money.”

That journey is accelerated. There’s a high level of acceleration happening right now.

Imagine you and I are interacting with a lead data scientist, and we both give that person a business problem to solve.

I’m sure nine out of 10 data scientists would show us 1,000 lines of code with area under the curve, ROC, confusion matrix, mean squared error, R-squared, p-value, and z-score.

We don’t want to see who the smartest person in the room is. That’s not the discussion.

The discussion is about how a business problem can be translated into an outcome and how data science is solving the problem.

For that, we need to have an interface where the business and technology teams meet, co-innovate, and solve the problem together.

That’s where the apps come into the picture.

In my journey at Kohler, this was the same problem we had with a demand-forecasting algorithm or customer segmentation.

We could not come to a business leader with 1,000 lines of code for a demo.

We started building lightweight Streamlit and React.js apps, put the model on top of the app, and delivered that app to the business.

Doug Barton:
That was building a bridge from problem to solution, with the model included in a way that supported translation and application.

Karthik Josyula:
That might not be the finished solution. It could be a simple prototype that gives the business an experience of how the technology is solving the problem or what the experience will be once it goes into production.

Doug Barton:
There is a lot of human-centered design and research that needs to happen at that interface.

If we’re getting back to adoption risk, you have to manage that risk, too.

You are then able to more quickly innovate your way to solutions in a risk-aware way. Those solutions will get deployed, used, and have an impact.

Karthik Josyula:
Now the business and technology teams are co-innovating together, and they’re building a product.

Doug Barton:
Your business and technology are combined, and now it becomes a product where technology and business are solving and building together.

You might use the term “experience” interchangeably there, such as a customer experience or product experience. Do you think about it that way?

Karthik Josyula:
When I say “product,” I mean the solution sitting on top of a capability that is being served or utilized by the business, a customer, or an end user.

Doug Barton:
Got it.

The Excitement Around Agent Bricks

Anything else? Do you want to mention Agent Bricks and the excitement around it?

Karthik Josyula:
With Mosaic, it’s the year of agents.

Doug Barton:
Were you aware?

It was hard to miss. I thought this was 2025, but it turned out to be the year of agents.

Karthik Josyula:
Yes, the year of agents.

Agent Bricks is another big game changer.

We were going through the generative AI journey, and now Agent Bricks has entered the equation.

It is mostly about solving enterprise work processes. How can we make enterprise work processes simpler and process flows more optimized?

Agent Bricks is a strong capability for that.

Imagine this from a Databricks perspective.

Now that you have your entire common data model residing in Databricks, you have enabled Unity Catalog.

Unity Catalog can also reference data from outside database systems. We can have Unity Catalog enabled with all the data points.

Now imagine how easy it is to build agents on top of governed, certified data.

Doug Barton:
Exactly. That’s an exciting future.

Let’s move on. I’m getting hot in the sun.

Karthik Josyula:
Me too.

Lightning Round: Insights and Reflections

Doug Barton:
Finally, let’s do a little lightning round.

We’ll have a chance to get back together and give the audience what they need in the future, but for the lightning round, I have three questions.

Actually, only three or four words here.

Favorite metric?

Karthik Josyula:
ROI.

Doug Barton:
Okay. ROI.

Karthik Josyula:
Return on investment.

You can put the ROI metric anywhere, including technology, people, and process.

Doug Barton:
Sometimes people think of ROI as a financial metric and therefore limited to that. No, it’s not.

Say more. Why do you think that?

Karthik Josyula:
I spend a lot of time with my team, and the return on that investment is trust.

I want them to trust me. I want them to believe in me as their leader.

For that, I need to spend more time and engage in deeper conversations, not just about work.

The time I invest and the care I show them return trust.

Doug Barton:
Thanks for deepening the idea.

There are returns for what we do. Something becomes available, something becomes possible, or something is enabled in return.

You build trust, and trust creates confidence in themselves and others.

Perhaps it’s curiosity that leads to the next big breakthrough. That’s the return, and you’re making the investment.

Thanks for sharing that.

What about an unaddressed risk or worry that you have?

Karthik Josyula:
I would say succession planning.

We are trying to build these unicorns on the team. There are definitely unicorns who have the attitude to say no to nothing. They take whatever comes in and want to deliver.

Technology is evolving so rapidly, and these unicorns are keeping pace with the technology.

They’re completing certifications, doing self-directed learning, looking for opportunities to work on these technologies, and solving business problems.

If there is a flight risk and the individual finds another opportunity, who is their successor? Who is going to be put in place?

As leaders and people leaders, we should start looking more deeply at identifying those unicorns and what their succession plan is.

That is one thing that is unaddressed right now.

Doug Barton:
I think there is a sense in which, as with any technological wave, there is a moment when you are dependent on finding the unicorn who has a unique combination of skills.

Over time, you want that to become an organizational capability and organizational muscle.

That’s why you’re thinking about succession. We don’t want our most important asset to walk out the door every night without a way to bring new people into that conversation and enable them.

Thanks for that.

Lastly, what is your candidate for something that is overhyped?

Karthik Josyula:
This might come as a surprise, but agentic AI.

When a new technology comes in and businesses are trying to evaluate where it fits in the organization, we need enough time to take a deep dive, analyze it, go through the technology journey, and then move to the next one.

With the rapid evolution of technology, I don’t think we have enough time. I wish we had 60 hours in a day instead of 24 so we could focus on multiple things.

Everything will be driven by the business. Business problems will set the direction of where we need to head as an organization, and technology will be a supporting buffer.

With the evolution of technology, technology teams are in a mixed situation where they’re handling the current business while also trying to catch up.

They don’t always know why they’re trying to catch up, and the business may not be ready.

I still want to see what happens with Agent Bricks.

There is also the concern about whether AI will replace humans. That is still not addressed.

The discussions and white papers look good on the table. But when it comes to actual execution, I want to see more use cases.

I want to take time. I don’t want to rush into these things, but I would love to learn and talk about all of them.

We need to immerse ourselves in the technology, understand it well, and evaluate what is good and what adds value for the business.

Doug Barton:
Being thoughtful and planful makes a lot of sense, especially in the world of agents.

In some respects, what makes them agents is their ability to plan and take action.

If you are going to give that responsibility to a nonhuman actor that might not have the context around your policies, procedures, and responsibilities, it could be problematic.

Yet I think the unlock might also be more experimentation. We can’t enter this future by only studying white papers.

We get to experience agentic AI in small applications today.

When I share a request with ChatGPT Team and use deep research, I can see the reasoning about how it’s going to research the question I gave it.

I’m enriched by that experience. It’s using reasoning that I might not have expertise in for the question I asked.

I think that enriches my work and informs what’s possible in the next generation of these systems.

I hope we take on experimentation.

That’s the end of our lightning round, Karthik.

Conference Announcements and Closing Remarks

Let me do some shout-outs here. This is new.

I want to offer a free Annual Conference pass to anyone who shares this highlight reel that will be posted on LinkedIn.

Karthik and I are just down the street from Monona Terrace. That will be the site of our Annual Conference on September 30.

We have terrific speakers. One of those speakers is none other than the author of the New York Times bestseller Co-Intelligence. His name is Ethan Mollick.

The subtitle is Living and Working with AI.

He just wrote a Substack article in the middle of May about making AI work.

Some of the things he talked about, including experimentation with the crowd, using a lab to bring inspirational ideas to fruition, and the role of leadership both to enable and direct us into the future, are really thoughtful.

He’s going to bring some of those ideas to our Annual Conference.

With that, let’s go walk around the Farmers’ Market.

Karthik Josyula:
Absolutely.

Doug Barton:
Perfect timing.

That’s all for now. Thank you. On, Wisconsin. Bye.

Karthik Josyula:
Cheers.

UWEBC Podcast Closing

If you want to keep the conversation going, check out the UWEBC website for more events, resources, and details on what’s coming next. Thanks for listening, and from the UWEBC, we’re glad you joined us.