FinOps X 2026 | Hunter Harris, Pegasystems | FinOps X 2026

Clips
More from FinOps X 2026

Hunter Harris

Director of Cloud FinOps

Pegasystems

Pegasystems uses FinOps foundation to navigate the unpredictable economics of AI

As AI spending surges, organizations are expanding their focus from AI cost optimization to value measurement. The trend is driving demand for deeper visibility into how AI investments impact productivity and business performance.One of the biggest changes AI introduced was unpredictability, as usage and costs can scale far beyond the patterns typically seen in traditional cloud environments. That volatility has forced organizations to rethink how they track and manage AI spending as adoption grows, according to Hunter Harris (pictured), director of cloud FinOps at Pegasystems Inc.“AI can more or less scale infinitely until the CSP runs out of capacity

play_circle_outline Reining in AI Spend: Classifying Experimentation, Internal Productivity, and Customer-Facing Costs

play_circle_outline Agentic Analytics: Leveraging FinOps Platforms and Data Models for Rapid AI Cost Integration

play_circle_outline Challenges of multi-cloud parity and tagged metadata for AI billing

Info
Transcript

Hunter Harris, Pegasystems | FinOps X 2026

Hunter Harris

Director of Cloud FinOps Pegasystems

In this interview from FinOps X 2026, Hunter Harris, director of cloud FinOps at Pegasystems, joins theCUBE's John Furrier and theCUBE Research's Paul Nashawaty to discuss how FinOps is evolving from cloud cost control into an enterprise-wide discipline for connecting AI infrastructure spend to measurable business outcomes. Harris details how Pegasystems built a granular cost attribution framework before the AI era — mapping spend down to individual microservices, cloud providers and products — giving the team the structural foundation needed to absorb AI wor... Read more

explore Keep Exploring

How should AI spending be classified and managed (for example: experimentation, internal productivity, and production/customer-facing products), and what challenges arise in budgeting and cost visibility? add

How did you determine why the observed activity (e.g., spending patterns or model usage) was happening? add

How are you tracking and managing AI-related cloud costs (e.g., tagging, monitoring inference/Bedrock charges) across multiple cloud environments, and what challenges remain to achieve a unified dataset and optimization processes? add

bolt Powered by CUBE AI

Hunter Harris, Pegasystems | FinOps X 2026

search

John Furrier

>> Welcome back over to theCUBE's livestream here in San Diego for FinOps X 2026 where we're unpacking how cost and value creation work and what does it translate to? Does it move the needle? Does it drive margin, revenue? Who was involved? Tokens are involved. AI is changing the game, but costs are exploding and now it's time to reign in the chaos. Hunter Harris is here, director of cloud FinOps for Pegasystems. Welcome to the Cube.

Hunter Harris

>> Thank you guys. Happy to be here.

John Furrier

>> Yeah. Let chaos rain and then rain in the chaos. It feels like Token Economics has done that and now it's like, okay, we see value creation happening. How do we instrument it? What's the cost? What did it tie back to?

Hunter Harris

>> Yeah, I think a big part of what we did is we were very big on KPIs before the AI economy came. So we were able to show, hey, what is our cost per customer, cost per environment? Map it back to every microservice that we had, what cloud provider was it coming from, what product, what EC2 instance type. So we really understood our cost, and can then break it down for our total cost per whatever, what each layer was. When AI came in, we had that structure already so we could fit it in and figure out, well, why is cost going up or down? What do we need to do to get to our target revenue? What are the things we need to optimize?

John Furrier

>> What was the learnings from that? Because you guys did the word cloud. You had to know all the mechanisms. Like you said, you had the layers nailed down. Okay, you're starting to do it. What changed with AI? What was the big factor and what did you have to adjust for?

Hunter Harris

>> Unpredictability a lot of the time. Just like generally speaking with cloud, it can scale up to a certain amount. AI can more or less scale infinitely until the CSP runs out of capacity. We haven't had any runaway like that, but the only thing that's predictable is that it has been very unpredictable in many ways. So with us, a lot of it was like, how do we classify our AI spend? Where is it going? Is it for internal productivity? Is it for our customers and products that we're building? Is it experimentation? We had to put it through that lens.

Paul Nashawaty

>> And so that's an interesting point. I've been asking this question throughout the day. When you mentioned experimentation, you mentioned productivity, optimization, understanding the cost, what it all means. Let's unpack that because that matters. The dollar is not equal to a dollar is not equal to a dollar.

Hunter Harris

>> Exactly.

Paul Nashawaty

>> And when you look at innovation and experimentation, that should be weighed probably less than production.

Hunter Harris

>> Yeah. A lot of it comes down to what are the company initiatives. Generally speaking, we want to be introducing AI to our products. We want to be unleashing value in our engineerings with agentic engineering skills, but understanding what we're doing in each one of those. So agentic engineering, we wanted to innovate. We gave all the tools available. Let's go at it. Initially, there weren't great budget constraints around it, so we're having to put user limits on. Maybe there weren't limits on what models you could use. You get different value out of different models. A lot of that is just bringing the visibility in. Executives are asking the right questions. We're all learning as we go, but being able to map it all back instead of saying, "Hey, the bill went from this to this." Classifying how it did that, showing what it was, why did it happen?

Paul Nashawaty

>> Absolutely. And when I think about what you're saying, there's a number of approaches to get there. There's a hundred roads to get to the same destination. And you picked one. And it could be you could do it yourself or you could partner with somebody. Now, you went down this road of figuring things out. How did you get to... Take us on that journey.

Hunter Harris

>> Yeah. A lot of it, we had the FinOps platform in place already that we built ourselves. We knew the data. So we already had most of the data there and we were seeing what it was happening, which accounts it was happening in, who owned those accounts. We could tell the original story, but more of it was why is it happening? What are you guys doing? What is the context of this? We knew which ones were our customer accounts, which ones were our R&D accounts, which VPs owned those R&D teams. So we had that, but then the inference of but why?

Paul Nashawaty

>> Yeah. So the red, yellow, green dashboard wasn't good enough.

Hunter Harris

>> It wasn't good enough. We knew the models, how they were changing and everything else, but then it also becomes, why are you using this model instead of that?

Paul Nashawaty

>> Yeah.

Hunter Harris

>> Again, that's like a lot of what this conference is about now is like, how do we get in control of that now? There's great people out here trying to figure that out and maybe we're getting to the period where build your own, We're not going to be able to keep pace with what the tools are doing.

Paul Nashawaty

>> I agree. So how did you figure out the why?

Hunter Harris

>> The why was us using AI too. Honestly, like training our own agents to say like, "Hey, we've been doing it ourselves. Let's train the agents on what all these tables mean in our data, and then be able to question it and build automatic dashboards, release that to analysts to unleash them, give it to engineers. They can start asking the questions from the data instead of asking us all the time because we're a rate limiting factor." So we're calling it agentic analytics is what we decided to do. We have agentic engineering side. Well, let's bring the data into it as well.

John Furrier

>> A lot of people are talking about the agents, the decisions they have to make we hear on the main stage here, model routing. You're starting to hear deep tech talk in a FinOps crowd. And of course, open source is highly technical, but you're getting at the business intersection now of business. So like decision making, that could be which Pareto curve to use. Where we were talking on X earlier about premium tokens. You start to see quality of service. These are systems problems.

Hunter Harris

>> They are, but I think it's a conversation problem. I've always said FinOps, they need to be a Rosetta Stone. You have to be able to translate everyone's language. You have to be able to talk to an engineer or a support engineer or a salesperson or a product engineer or an ops person and get everyone on the same page. That's why I think we've been successful. It's my previous life, I did all those things. I ran cloud for a FinTech company, and I knew how to talk about these things in context to drive business value. What we did too is like we're not just talking about the cost per environment. We mapped it all back to revenue at the contract level, so we could see really like what is the margin of every single customer, and then what is the overall contribution margin of our cloud?

John Furrier

>> So you basically built a contextual engine on your system to do that translation.

Hunter Harris

>> Across everything-

John Furrier

>> So that pays dividend. You did that before AI came in.

Hunter Harris

>> Exactly. So we knew when the AI spend started coming in, yeah we have to tag things differently. Not all AI can be tagged. But we had at least the inference of everything else to say, "Okay, oh, here's bedrock charges coming in." Well, how's that coming in? What is it actually doing? How concerned do we need to be about this now and what is the scale and the future?

Paul Nashawaty

>> How much of focus applied to this kind of equation?

Hunter Harris

>> So focus is something we still want to implement, but truthfully, we're across many clouds. We're now using the AWS European Sovereign Cloud and everything. We're waiting for some of the functionality to catch up.

John Furrier

>> Gotcha.

Hunter Harris

>> We still want to do it. We're planning to do it all, but we need it all to be on par. We need that parity, because it all flows into all of our processes, whether it's rate optimization, use optimization, allocation. I would love to have one data set, one schema, to push it all through. It would be great.

John Furrier

>> Okay. So I have to ask you like this is like a real time riff consulting session. Say I'm a CFO, but I'm an operational oriented because there's two types of CFOs we're seeing in our research.

Hunter Harris

>> Ours is an operational CFO.

John Furrier

>> It's like, "Oh, I am a guardian of the books on one hand. That's the way we do it. Here's your budget." Then you see more COO behavior. They want real time. So what comes out of this community? Because right now we're still in the emerging stages of multiple adjacencies converging. But FinOps I think will ultimately point to resulting benefits to the CFO and team to manage their business. What would you be advice to me? What do I do?

Hunter Harris

>> Understand your data. You have to connect all of this into a data model. It's not just cloud cost data, it's your operational data, it's your support data, it's your product data, it's your bug data, it's your revenue. If you can actually have all of that come together, you can tell real stories about how the business is performing, and then make much more accurate forecast and budget decisions and predict. Sales forecast is very difficult. It's very, very difficult to forecast revenue, but if you have trust in the underlying operational metrics, you can set more guardrails around, "Hey, this is where the confines of our upper bound and lower bounds of everything."

John Furrier

>> So operational systems become a critical plumbing.

Hunter Harris

>> They do.

John Furrier

>> Even in forecasting, like say sales.

Hunter Harris

>> It is, yes.

John Furrier

>> So you can actually apply AI to that. Okay. So how are you guys thinking about AI in your platform and how are your customers thinking about AI as an augmentation?

Hunter Harris

>> Yeah. Historically Pega, it's a rules engine, it's a workflow, everything else. You've worked with some of the stuff we've done with Blueprint and other things. It initially was a sales tool. Customers can come in, describe an app that they need. And it would generate it on the fly, which is awesome. It's a great way to sell. Now we're taking it and building an end of a platform. So in existing customers of ours, if they're on the latest version, they have all these workflows and everything else. They can now use AI to augment those workflows, change them and everything else. That's huge for our customers, taking out the complexity of consulting and everything else.

John Furrier

>> All right. What's the coolest thing you've seen come out of the work you guys done?

Hunter Harris

>> Honestly, being able to demo our system on the fly is huge. If you think of any salesperson, you can go and I could say, "Oh, I'm in San Diego. I want to build a surf shop and rental of surfboards and everything else, supply chain, marketing, everything." Be able to describe what you want and have AI go and build the application for you, build a call center to take in routes, have a mobile app, all of it on the fly in seconds. That's crazy.

John Furrier

>> Yeah, that's insane.

Hunter Harris

>> It's awesome.

John Furrier

>> Who are you disrupting with that? Because that means basically the supplier network, the service providers all collapse.

Hunter Harris

>> It's modernization is what it is. All these legacy apps that are-

Paul Nashawaty

>> I was going to say, it's an ecosystem too.

Hunter Harris

>> It's an ecosystem. Having a platform to build all of that around where you can just describe it, you can take old schemantics of what those workflows were, feed them in as an attachment and have it build your app on demand and have something to demo and just go. That's crazy.

Paul Nashawaty

>> Yeah, it's fast.

Hunter Harris

>> Yeah.

John Furrier

>> Well, I'm psyched. What do you think about the show here? Obviously it's a transitional time. We got the new foundation for tokenomics, which is a nice separation you talked about earlier, but what's the future look like? Because I really see the FinOps becoming much bigger and broader. Again, the C-suite will be impacted. Not that the JR should call on CFOs to come here, but this body of community will generate value that will spin directly to the doorstep of the CFO, the CHRO, CISOs and CIO.

Hunter Harris

>> Yeah. Originally it was all cloud value is what it was. And then it was technology value and obviously AI is technology. It's been great for us because we had the framework, people trusted us already as AI has gone, I'm not going to say off of rails, but as it has grown, it's like, well, who do we trust with it? We have the foundation there. The business trusts us already. That's natural. I think a lot of what we're trying to do now is take it beyond just the concept of cost but drive into value. Which is again, a huge theme here. Taking it, being able to talk to our chief product officer, look at, hey, what is our output, our engineering teams? What did it cost? What is the cost per bug and resolution? How much code is being checked in? All of that is there. We have a foundation to do that now .

John Furrier

>> Hunter, final question for you. What's changed the most in the past year in this community?

Hunter Harris

>> I think expectation, which is great. I think it is for us, again, we had the data and everything else, but expectation that it will be immediately available, adaptable, et cetera, we had that. It's just the speed at which we do it. And I think it's confusing sometimes around, well, how should we do it? Do we trust AI to do everything for us? Everyone's still figuring that out. What do you offload to an agent versus what do you do yourself?

Paul Nashawaty

>> Yeah. And it's a no light trust thing, right? You have to trust it that's going to do the right thing until then there's a human in the loop.

Hunter Harris

>> Yeah. And again, the changing in going to token-based consumption versus a per user thing, it's a huge shift in the past few weeks for everyone.

John Furrier

>> Yeah. It's a great time to be an engineer, great time to be an AI builder, great time to be interested in real time data.

Hunter Harris

>> Yes.

John Furrier

>> Hunter, thanks for coming on theCUBE.

Paul Nashawaty

>> Thank you.

Hunter Harris

>> It was a pleasure guys.

John Furrier

>> Appreciate it.

Hunter Harris

>> Thank you.

John Furrier

>> All right. I'm John Furrier here with theCUBE and Paul Nashawaty, doing our part here at FinOps on the ground, getting all that, unpacking. This is the future. The cost equation is instrumental because the value creation is happening. ROI, all these calculations driving the value proposition needs to be nailed down. That's what they're doing here. We're doing our part to bring you the content. Thanks for watching.