Agentic AI in Action: Turning Data into Outcomes

Tom Soderstrom:
Welcome to the Executive Insights podcast, where we talk about insights that we have gotten from our customers worldwide. I am Tom Soderstrom, I'm an enterprise strategist, and with me is ...

Miriam McLemore:
Miriam McLemore, also an enterprise strategist.

Tom Soderstrom:
We're part of a group that has led transformation for our respective companies or agencies. Miriam, very impressively, did...

Miriam McLemore:
Coca-Cola.

Tom Soderstrom:
... Coca-Cola, and I did Jet Propulsion Laboratory in NASA.

Miriam McLemore:
Just NASA. No big deal.

Tom Soderstrom:
Just the cloud thing. If you are watching this, and we think you will get something out of it, especially if you're a business leader or if you're in IT, worrying about what's going to happen with all this data, or if you're just trying to figure out how to rise through the organization and probably hit the biggest thing that's happening, which is data, and what can we do with it. Now, Gartner studies are saying most people are not happy with their insights of data.

Tom Soderstrom:
So when I ask the executive leaders, I say, "What do you think about data," I hear, "Intuitively, I know the data is the lifeblood of the organization, but I'm overwhelmed. I don't know what to do. How much should I spend? How much can I spend? What should I do?" Or, "I have all this data, some I know of and some I don't. Can I monetize it?" Or, "Is the quality of my data good enough? Should I just wait for it to be perfect?" Or, "How much is it growing?" I'll come back to how much it's growing. It's pretty astronomical. What are some quotes you are hearing from your customers?

Miriam McLemore:
The issue that I've seen with customers is up 'til now we've relied on structured data, because that's what fit in our tables. That's what we could calculate. That's what we could populate to dashboards reliably, with accuracy, but there's this huge and growing world of unstructured data that customers are looking for access to. The world of agentic gives you the possibility to leverage that unstructured data and integrate it with the structured data, but those are new metrics for business leaders. This is a different way of assessing your organization, the place you play in, the opportunities and outcomes that you can drive. The questions I'm getting from customers is, "Where do I start? I'm overwhelmed. I have a bunch of this structured stuff. There's all of this unstructured. Where do I begin?"

Tom Soderstrom:
Yeah, it's paralyzing for them.

Miriam McLemore:
It is.

Tom Soderstrom:
I thought I would pull some actual quotes of data.

Miriam McLemore:
Yeah.

Tom Soderstrom:
I did a podcast with Mai-Lan Tomsen Bukovec, who owns data at AWS.

Miriam McLemore:
Yeah, of course.

Tom Soderstrom:
She gave me some interesting stats about S3, their primary data source. It has 1 million data lakes on S3, holds 1 trillion objects, has many, many exabytes. Averages 150 million requests per second. Has thousands of customers that store more than one petabyte, several that store more than an exabyte. For a business executive, this is exciting. "Look at all the stuff I own." For an IT person, this is horrifying. "How do I manage it?" I think the growth of data ... and I think you said it well ... where do I start? Just start. And 10% is structured, the rest is unstructured. With agentic AI and generative AI, you can take advantage of it. I think that's something we're going to talk about, successes and failures going forward.

Miriam McLemore:
Well, it's interesting, Tom. At least I'm finding that financial services, insurance, and in some spaces healthcare, are jumping in faster, because their data has had to be clean. They've spent a lot of time on accuracy, integrity of their data. Other organizations have a lot of data silos and inconsistencies in their data, so there's data cleansing, there's data governance, that they have to put in place to leverage this new technology.

Tom Soderstrom:
Yeah. It mirrors what I'm hearing. We both deal with customers across the world, large customers, and anybody who's really regulated, we see that as a disadvantage, but it's not. It's an advantage, because they already know how to deal with their data. The public sector, where I grew up, certainly does. I thought maybe what we would do is just talk about some of our customers that we've seen done it incorrectly, and some that we can learn from that have done it correctly. Then maybe we'll cover some actual customer examples with names.

Miriam McLemore:
Yeah.

Tom Soderstrom:
Do you dare?

Miriam McLemore:
Sure.

Tom Soderstrom:
All right. Unsuccessful, they create too much friction. I see this all the time. The person who owns the data ... quote “owns the data” ... doesn't want to share it.

Miriam McLemore:
Yep.

Tom Soderstrom:
If you want my data, you have to fill out a form. In fact, two or three forms. We want everybody to use the data, so what's the answer? Flip the script.

Miriam McLemore:
Yes.

Tom Soderstrom:
If I need to protect my data, I fill out a form. It's an easy thing to do. For instance, at one big space agency, we noticed they spent as much money protecting the cafeteria menu as spacecraft uplink commands, but you don't know that until you look, until you see.

Miriam McLemore:
It is fascinating, because we were in a world where data was power. If you own the data, you control the data. I had that with the functions that I was interacting with - finance were the only people who could see the finance data. The people in science were the only people that could see that data. These silos perpetuated themselves, because it became a culture within those organizations that it's not safe to share. I love this concept of flipping the script and talking about the value, and data is an organizational asset. It's not a functional asset.

Tom Soderstrom:
Exactly. We had talked about coming up with an organizational data mesh where it's really about incentives. It's about the people who want to use the data, really want to use it.

Miriam McLemore:
Yep.

Tom Soderstrom:
The ones who produce the data want people to use it. So why doesn't it happen? You need somebody in the middle that just makes it happen, and make it incentivized. Remove the friction.

Miriam McLemore:
Sometimes the data governance can become the friction.

Tom Soderstrom:
Yes.

Miriam McLemore:
You also have to do that the right way, because I agree with you. There are producers of data and consumers of data, but in the middle, you do need the ability to manage that data and govern the data and have the right security, the right identity and access, management controls. All of that is absolutely important. We're not saying it needs to be a free-for-all.

Tom Soderstrom:
Right.

Miriam McLemore:
We're saying that you've got to bring insights together to drive outcomes. It's interesting. I had the great opportunity to attend a Formula 1 event just recently.

They are such an amazing example of being data-driven as teams, as drivers, and as the overall F1 organization. What I appreciated about them, which is a tenet that I believe many organizations can take, is nothing, no sensor, no data insight, is added to a car or added to the screen, because these are split-second or split-into-millisecond  decisions that need to be made. They ruthlessly prioritize what insight matters, what actually drives a business outcome, because of course, in Formula 1, speed matters.

Tom Soderstrom:
Yeah. In fact, of all the executives I've talked to this year, and I'm sure you too, the number one priority is speed.

Miriam McLemore:
Has to be.

Tom Soderstrom:
Speed to market, speed to profitability, speed to compliance, speed to skills. And the data can help all that. It's an exciting, exciting time. The other thing I'm seeing that companies do wrong is they're trying to be perfect. Like your Formula 1 example, if you were just theoretical and never tested it, you wouldn't know. Start with a business use case, and try it and try it and iterate, and expect the first version to be the worst and the most expensive, and then you can move forward. Data really enables all of these other things, and we're going to talk about agentic very shortly.

Miriam McLemore:
Yeah, but that culture of experimentation, I just don't want to leave that important point, because as we lean into integrating more data, you have to have a culture in your organization that allows for us to learn and take those improvements. That's what agentic does. These agents are self-improving agents. They'll see what the path was, and they'll start to learn is there a better path? We have to also have that culture inside our organizations.

Tom Soderstrom:
Yes. Again with the incentives, if you ... since like that's what I've wanted ... work backwards? What's the end result that I want? How can I get there, and how can my data help me? Lonely Planet took their petabytes of data and they organized it. They used generative AI and they came up with this virtual travel agent.

Miriam McLemore:
Yes.

Tom Soderstrom:
Kone Elevators, a Finnish company, hundreds of thousands of elevators. The person goes out to fix it and they have all the information at their fingertip. I think one of the things that we see through all of these is that everything is data. Code is data. Amazon changed, upgraded all the Java. Sounds boring. It's expensive. 3,500 work days were saved, $250 million saved, by using generative AI and agents and treating the CODIS data. It's a fascinating new world. What's your favorite agentic example?

Miriam McLemore:
Well, I actually like the one you just mentioned with Lonely Planet, maybe because I travel a lot. What they've done is really get clear on the traveler's need. What does the traveler need in a travel agent? What's the help that they're asking for? Where should I stay? Where should I eat? How do you make my life simple? What's the Uber equivalent in this country? Help me travel, and make that simple for me and use data. That's one of the things that I love about the Lonely Planet example.

Tom Soderstrom:
Another example that I really like is NASA data is now in AWS Data Exchange, and it can now help predict floods in Australia, droughts in Africa, because everybody can access it. A professor at University of Sydney put all of her genomics data on Open Data Exchange, and a researcher in Sweden solved how to help save the koala. It's just this global ...

Miriam McLemore:
I love that.

Tom Soderstrom:
... help. If you were going to give advice, to the future leaders who now is going to manage this, what would you do? What would you say?

Miriam McLemore:
Yeah. I had roles where I was the data czar, and because I worked in a global business, we were always trying to bring data together, but we were doing it for the purposes of reporting, right? Because from a finance standpoint, we had to report consistently. That was a reasonably good reason, but it didn't drive business. What I would encourage new leaders today is to start with what are you trying to accomplish. What is the business need you're trying to drive? I mentioned F1, so bringing down a pit stop from what it was, I think in the '50s, 67 seconds. A pit stop can now be accomplished in 1.8 seconds.

Tom Soderstrom:
That's amazing.

Miriam McLemore:
Just anything possible, but it is about taking out all the things that don't matter and only focusing on the things that do, and then coordinating them to the millisecond. How do we do that in business? Leaders being inspired to drive business outcomes and constantly improve. I think it's an exciting world.

Tom Soderstrom:
Yeah, it is an exciting world.

Miriam McLemore:
To be a leader.

Tom Soderstrom:
I really agree with what you said, focus on the things that matter. That goes with data too. Don't worry about it. Just use the data that you can use to create this outcome. Work backwards and focus on that, and the rest will come. These future leaders have to show results all the time. That comes from a culture of experimentation and just iterating. I would say, for the data owners out there that feel they are in control of data, use that to drive these new generative AI and agentic results. Show business outcome.

Miriam McLemore:
Yes. To your point about leadership, we've got to start at the top, get the leaders engaged, but we've got to come up from the bottom too and upskill, so that we meet in the middle as an integrated team.

Tom Soderstrom:
Thank you very much. Thanks for teaching me yet again. I appreciate it.

Miriam McLemore:
I always enjoy my conversations with you, Tom.

Tom Soderstrom:
Thank you.

Listen now

Listen now

Listen now