The Real Cost Of Enterprise AI

AI isn’t magic, and it definitely isn’t free. We sit down with Ken from Pega Systems to get brutally practical about the economics of enterprise AI: why token costs are a symptom, why infrastructure spend is so high, and how “murky ROI” happens when companies deploy AI for novelty instead of measurable business value.

From Ken’s perspective as a former CFO and current COO, the best mental model is surprisingly simple: treat AI like a utility. If electricity has taught us anything, it’s that the winners don’t just consume more, they manage consumption better. We talk about how to reduce waste, how to avoid paying for frontier-model overkill, and why boards and finance teams are starting to demand tokenomics tied to outcomes. We also dig into a provocative corner of the market: incentives that can turn the AI ecosystem into a circular hype machine unless leaders insist on real examples and hard metrics.

We then shift to what this means inside large organizations. Agentic AI can accelerate judgment-heavy work in finance, legal, HR, and marketing, while deterministic workflows still anchor reliability in core operations. Finally, Ken shares career advice for the next generation: as execution gets automated, the premium rises on strategy, product management, and validation skills, plus the curiosity to keep learning as roles evolve.

If you care about enterprise AI ROI, workflow automation, and the real operating model behind digital transformation, hit play. Subscribe, share this with a colleague, and leave a review with the metric you think will prove AI is paying off.

Transcript

Welcome And What Pega Does

SPEAKER_00 0:11

I'm really excited to speak with Ken at Pega Systems. Ken, how are you?

SPEAKER_01 0:14

I'm great. Great. Happy to talk to you.

SPEAKER_00 0:16

Yeah, really excited. How do you describe Pega Systems these days for folks who might be new to Pega World?

SPEAKER_01 0:23

I typically still do use the word workflow because I think it's a I think it's a commonly understood thing that you're trying to process, you know, scale amounts of transactions and work. And I think that is fundamental to who we are. I think the the other the other piece of this is really how much we've been helping our clients digitally transform and reimagine their businesses. I mean, many of these businesses that have been around for decades, um the business has changed a lot, but the technology really doesn't keep up. And I think that we're trying to help uh help not only automate and build efficiencies and and leverage AI, but really help help our clients reimagine like who they actually are and who their client is and how that experience uh you know should uh you know should should exist.

SPEAKER_00 1:08

Yeah, it's been on full display, really impressive. How

From CFO To COO In Cloud

SPEAKER_00 1:11

do you describe your role at Pega Systems and how has that evolved over the years?

SPEAKER_01 1:15

Yeah, I I I came into Pega Systems, so uh you know, amazingly, uh I will be hitting my 10-year anniversary of Pega Systems. So it's been so I so I've been here long enough that uh you know I I kind of um you know I I think the formation of our strategy um is uh you know I would I would certainly take full ownership for the good and the bad of that. Um but I started as the CFO. Um and um, you know, I I I helped to to really drive the business model transition of moving to the cloud. And um and I think what uh you know uh what I was rewarded with for that was I became COO and got to own the cloud. And so and so uh so I I view my role now as as much as really kind of a broad leader across the organization. So I have areas that direct report to me. But if you think, if I think about what I do, I I really try to make the whole thing work, right? Across engineering, across sales, across uh our people function areas that don't directly report into me, um, and also the areas that do. So uh, you know, I just try to try to help our board and and Alan and our leadership team, you know, just make sure that we're you know doing the things we need to do to be successful with our clients.

AI Economics And The Utility Analogy

SPEAKER_00 2:27

I'm really interested in your financial background, in particular because the economics and finance behind AI is a top uh topic today. As and what's your take as you know, CFO, former CFO on the economics, the the headline $40 billion invested so far in AI, and the return is yet to be evidenced. Um, what does that mean to you as as a CFO?

SPEAKER_01 2:51

Well, so what's what's what's really interesting is if you if you start from the starting point of 1.5 trillion invested to build the capabilities for AI, you know, 40 billion, use your number, 40 plus billion dollars of of of of spend attributed to that, um, ROI elusive, right? I mean, is is probably uh murky. Yeah, murky murky is probably fair. So I I think I think the the question about the cost of tokens or the cost of AI or the cost of compute, however you want to characterize that, shouldn't surprise us. If you're gonna spend a trillion and a half dollars, and to be honest with you, a lot of the technology doesn't have 10-year life. I mean, some of the technology is 14, 16 months where you have to replace some of this. And I think with the with some of the reasoning capabilities and the frontier models, it's that much more expensive. It shouldn't surprise us that it's gonna cost a lot to leverage this infrastructure. So, and that's not necessarily a problem, that part of it. When you start to use AI for things that you don't really have a clear understanding of ROI, or you start to use AI for things that quite frankly are novelties or things that are done, then you there's really a lot of waste that is at risk. And I think that's where that we're at this, we're at this almost this point where we have to start to really be thoughtful about how do you use AI, use it for the right things, because it is going to be expensive, but you'll get value if you use it for the right things. And also using the models that you use and picking the right models, because naturally the most expensive model may not be, it might be overkill, you might get diminishing returns. So as a finance kind of as a finance person, I look at it like unsurprising that we have to, that the vendors have to get a return for this. And that's not the problem. The problem is the unnecessary use of AI. And the example that I use is I think of AI as somewhat analogous to a utility. It's like electric, right? That's what I think it will become. And what have we done over the years around utilities? We've insulated our homes, we've created smart thermostats that know how to monitor the temperature in your house. You don't leave your lights on. There's sensors, right? Office buildings don't turn air conditioning on on the weekends because nobody's there. We do all these things because we're really conscious to why would you waste money? Why would you waste you? And because there's because there's a significant cost to building these infrastructures. So I view this analogy as it is, AI is somewhat of a utility. And then we manage utilities. And so I think right now we're in this phase where there's a lot of trial and error. And I think that's okay because you're you have to be educated. But I do feel like we've hit the point where companies are gonna have some very difficult decisions to make. The last point I would make on this, and I think this is this is one that is um, you know, somewhat, I think it's very visible, but but maybe a slightly uh controversial topic. Much of the AI ecosystem is somewhat circular, right? You have you have you have a company that's selling a product but lending the money to the company or investing so they can have the money to buy your product to sell to another company that actually is going to get compensated for an IPO on that exact same company. So, you know, I do think we just need to be conscious of any perverse incentives that could exist in the system to so that's uh those are just things like, you know, this is not, it's not um, it's not unusual for that to be the case because, you know, financial incentives will drive uh, you know, different creativity around the transactions. But I think the fundamental thing is this is a powerful technology. It is expensive to run and should be used for the right purposes. And if we do that, if we use them for the right purposes, I really believe the value's there. It's using them for the wrong ones, is where the uh efficiency hits.

SPEAKER_00 6:44

Yeah, so well said. And we've moved from you know token maxing to like tokenomics. And so the CFO and board are now firmly looking at that. Um,

Outcome Based Pricing For Real Value

SPEAKER_00 6:55

and in terms of incentives, you're moving to more of an outcome-based model. What does that look like from your perspective? Also internally, how do you think about outcomes?

SPEAKER_01 7:04

So Pega has been um has been focused on this for 10 to 15 years previous to AI, where what we thought of with our clients is if we had a seat-based license, and one of the things we could help our clients with was to allow them to scale efficiencies with their teams. A light a seat-based license doesn't match the value play between the vendor, Pega, and the and customer. So what we decided years ago was to think about a usage metric, you know, a usage or an outcome metric, which is how many claims do we process, how many loans are originated, how much work does the system do? And we we felt like that was a clear connection to the automation. It was almost a, it was a it was essentially a proxy for the amount of automation that the clients were getting because the system was doing the work. So that now is even more relevant with AI, where AI is a sort of an employee, right? You could count that as a user. You have customers doing self-service, that's a user. You have your own employees. If you have the user model, it's really hard to gauge how much should you charge for all those different types of interaction. So it really comes back to what outcomes are you driving? Are you, you know, and you're getting work done. That's the way we think about it. The outcomes we're driving is to automate work. Therefore, we license our clients in how much that work is automated. That's kind of the connection point.

SPEAKER_00 8:26

Brilliant.

Where Agentic AI Helps Most

SPEAKER_00 8:27

And uh agentic is all the rage. Uh definitely. And you've got what, a thousand people across all the different functions that you help manage. Um, what changes for them internally and for your clients with these very large organizations when agents start to hit the streets?

SPEAKER_01 8:45

So there's some the way I think about it is that everywhere where a human being does a task that requires some level of judgment, decision, analysis, it is river for AI to optimize that. In a place like workflow, where you actually build workflow to take away the judgment of a person, AI isn't. We don't believe AI is relevant or as relevant in that deterministic workflow. When you go back to the judgment aspect, so things like legal and finance, accounting, HR, marketing, these please, all these areas where there's lots of human beings making judgment calls or analysis of information, and they can't do that as fast as the model they can when enabled by the models. So I think that's a great use. Those are great uses for AI. I mean, just even the example that Alan used on stage around the way that you learned was you actually, you know, you pulled an encyclopedia out and you found out what chapter and you read it versus now how would we learn, right? You ask an agent a question and it gives you an answer. Now, it doesn't mean the answer is 100% right, but it's a starting point for you to on that journey of learning and it's a much more efficient one. So I think that's that's kind of the way we see AI enabling. Whether that reduces the amount of people that organizations have, or it dramatically changes the profile and the actual description of the job, so to speak. Um, I think that's still under debate. It certainly is going to change the description, the job description of what people do. The question is, do you get more done with the same amount of people, or do you get the same amount done with a lesser number of people? Those are, those are probably situation dependent. But um for a lot of our functions, we view that there's a backlog of work that we wanted to get done that we weren't able to get done. And if we can actually use AI to speed up the productivity of that, we can get so much more done with the people that we have.

SPEAKER_00 10:41

I love that. What a what a great take.

Explaining AI Reality To Investors

SPEAKER_00 10:43

And you have the joy of of uh leading investor calls and answering investor questions. Uh lucky you. Um, what do they keep asking you about AI that they that you wish maybe they asked differently? Or how could you educate them in in a different way than we're seeing today, which is based on headlines in the Wall Street Journal, basically?

SPEAKER_01 11:03

So the so um I have I have um often said that um there's there's there's two traits about investors that if you understand, you can better communicate to them. The first trait is investors are very intelligent, typically because of their educational backgrounds, because of how much they see, their analytical skills. Second thing is they're very busy slash uh lazy, right? And that's not like that's not a personal, that's not a that's not a personal attack. It just means you have a lot going on. You know, I mean the average cell sign analyst covers 30 to 40 companies. It's impossible for you to be an expert on all of that. So I think first there's a high, there's typically a high level of the ability to understand and process things in intellect capacity. And there's also just, I only have so much time, so I'm gonna take shortcuts to figure out things. So for us, it's really important to try to make that message clear. When we talk to our financial investors, we talked about we want to be a rule of 40 company, we want to move to the cloud, we wanted cloud choice, here's why. Those are things that you have to just continue to nurture. Now we're in a world where investors are getting so much noise and so fast the noise, you know, that what happens with the next release? What is that, what are AI companies saying they're doing? How much is the spend? You know, is it really what we what we hear or not? These companies aren't public yet, so we don't really know what what you know what the what the uh economics look like. We do know how much is being spent on on infrastructure, and then they hear the narratives around the SASPocalypse or you know, uh SASACR, someone called it once. I mean, all these, and and it there's there's a lot of stuff that like can partially resonate, and there's stuff that doesn't make sense. And I think the part that that investors, I think, have really gravitated onto is we know there's a place for AI. We understand the value, we've seen it ourselves. The the concept of deterministic workflows has really resonated with investors because they're talking to clients. And those companies are saying, I'm not gonna AI my ERP system, right? I mean, like that. So I think they understand that there's a difference, but because of so much of the noise, I think it's it's still creating this gap of like, what is real? What is reality? And so I think our job as a participant in that conversation is to continue to use examples, to use client examples, to have even encourage our clients to talk about what's real and what isn't real. And then naturally to use that to explain our roadmap and how and how we're enabling our clients. And then we have to execute, obviously. So I think it is it is a it is a challenging time because it's disruptive, it's confusing, but I think we just have to anchor on those two things. Investors and analysts have a high capacity to understand and a limited time to be able to do it. And so we just got to try to continue to simplify that message.

SPEAKER_00 14:00

Love it. Simplify, simplify. It's always good. Let's talk

Career Advice In An AI World

SPEAKER_00 14:03

about the people side. You're you're known for building a culture of uh engagement and uh collaboration. And you know, the energy was here at Pega World. Uh, really impressive to see. You came up through the CPA side, the financial side of the house. Lots of questions about young people on what this means. Uh, should I go into a field like accounting? Should I go into computer science even? Um, lots of debates, including with my kid. Um, how does that shape what how you think about uh AI and your advice to this next generation of employees and uh and and workers as they think about their career path?

SPEAKER_01 14:41

That's a great question. So um I too, I have I have two uh two of my children are in college right now, and so I have these conversations real time. Um the way I see the world changing with AI is if you think about there's a there's really kind of a four, in my view, there's a four-stage approach to how kind of innovation happens. There's a strategy stage, which is where are we going? What can you know, what problem are we trying to solve? There's a pro, I would call it product manager or product management stage, which then takes the strategy and formulates it into an actionable set of here's how we're going to do it. Then there's the actual work that's done. We'll call that coding for right now, right? And then there's at the back end the validation step, right? And so what I think really will dramatically change is that is that execution stage will get automated more. It'll thin the layers of management and it'll put a heavy, heavy premium on people that can see what's happening and understand what's important, see the problem. That's more entrepreneurial skills. The product manager function of how to translate that and communicate that now to models potentially, or to you know, to a combination of models and people. And then the the validation or the skeptic side of the back end is I want to make sure this actually works. What was it supposed to be? So I think I've what I've been telling my kids is one, you have to be curious, you have to understand and be observant around options that people have, the companies have, that, and then you got to make sure that you understand like how does the problem get solved? And then you and the other part is you can be a big contributor into ensuring that what was supposed to be built was actually built. So job, though, roles like accounting and consultants and engineers, they will be changed dramatically. But if you are a curious person that loves to learn, that can focus on how to formulate an actionable way to get something done, or someone that's really good at validating that it actually works. Those are somewhat two different skills. AI, you'll you'll be you'll thrive in that environment because that's what companies are gonna look for. They're gonna look for people that can explain to the models and also manage them and on the back end, make sure they did exactly what it was that we wanted them to build.

SPEAKER_00 16:58

That's a great

The Metrics That Prove AI Pays

SPEAKER_00 16:59

take. Uh so final question. We talked about the murky ROI for AI in the enterprise. Uh let's look at this time next year at Pega World 2027. Um, what's the financial metric you think you'll be looking at to see if and when AI is paying off? What's uh the leading indicator do you think?

SPEAKER_01 17:18

I I think I think that we're as a as a as a business community, I think we're going to have to have clear ROI examples. A year from now, if we are still um wading in the waters of I don't really know what value I'm getting, you you won't survive. So I think between now and probably the next quarter or two, I think you're gonna have a lot more focus on, you know, there's by the way, it's completely fine to do experimental. So if you just say I don't care about ROI because I'm just trying to experiment, but that's not operational, that's not scale. But if you're actually building or transforming your business, you have to have a direction that you're going. You have to understand what what outcome. And then the out, if you understand the outcome, you could start to measure the ROI. So I think most companies are going to be really anchoring in or laser focused on. I've actually implemented AI here, and this is actually the outcome that I got. Could be could be innovation, could be uh customer uh satisfaction, could be cost or economics, could be where you sit in the value chain, how fast you can transform and innovate, take market share. Lots of different ways to get there.

SPEAKER_00 18:28

What an exciting time. It's gonna be a great thing to watch. It is. Thanks so much. Awesome. Thanks everyone for listening. Watching. Thank you.