In this episode
Ken Stillwell, Chief Operating Officer and Chief Financial Officer at Pega, explains where enterprise AI spending gets wasted and how he applies financial discipline to it. He tells host Greg Kihlström that the biggest cost comes from using AI for routine, rules-based work that a workflow already handles cheaply at scale. He describes how Pega routes internal AI requests through a small steering committee, why a proxy outcome such as a faster quarter close can stand in for a full ROI model, and why leaders should be rewarded for shutting down AI projects that are not delivering.
Key takeaways
- The biggest AI cost is using it for routine work. Stillwell says the largest cost hit comes from people using AI for structured, rules-based or workflow-based tasks, where a workflow, once set up, can run billions of transactions very cheaply.
- Use AI for what a workflow cannot do. Summarizing a set of conversations into two paragraphs is a job for AI. Deciding who to reply to on an email, when the sender’s name is already in front of you, is not.
- Pega routes AI requests through a small steering committee. It is led by the company’s engineering AI lead and its CIO and represents every function. Stillwell compares the alternative to the early cloud era, when teams bought AWS instances on credit cards.
- Pega started with exposure and moved to discipline. Early on, the company lowered barriers so people would use AI, with governors on what they could do so cost did not get out of hand. It has since moved to being specific about which problems it is solving.
- Pilots get out of hand when nobody thinks like a product manager. A collections manager should point AI at the collections team’s problems, and sales operations at the selling team’s, rather than building interesting things that solve nothing.
- A pilot costs more than tokens. Stillwell counts the people and management focus shifted onto a pilot as part of what it costs.
- A proxy outcome can stand in for a hard ROI calculation. If AI could take two days out of the quarter close, or bring cash in five to seven days faster, he says that is enough direction to start without a complicated model.
- Shutting down a project should be rewarded. Stillwell says boards and executives should credit people who acknowledge AI is not helping in one place and shift resources to another, instead of treating it as failure.
- AI will become as assumed as email, but not yet. He says email has had a decade of security hardening and AI has probably not had a year.
Chapters
- 00:58 — Intro: what sits under the “AI” line in your budget
- 03:05 — Ken Stillwell’s role as CFO and COO at Pega
- 03:54 — What Pega does: regulated workflows at scale
- 05:40 — Why Pega routes AI requests through one steering committee
- 07:25 — Exposure first, discipline second: Pega’s two phases of AI adoption
- 09:47 — The biggest AI cost: using it for routine, rules-based work
- 15:07 — Keeping pilots from fragmenting the organization: think like a product manager
- 19:30 — Proxy outcomes that stand in for a hard ROI number
- 22:42 — Shutting down an AI project, and why leaders should reward it
- 25:15 — When AI becomes as assumed as email
- 27:34 — One year out: commodity models versus frontier models
- 28:21 — How Ken stays agile
Why AI requests at Pega go through one small committee
Stillwell says Pega centralized anything related to AI from the start. Two people anchor it: the engineering leader who runs Pega’s AI initiatives and the CIO, who manages vendors and connects the data. Requests are channeled through a small AI steering committee that represents all functions. His reason is what happened with cloud, when people used credit cards to stand up their own AWS instances until the company had to pull it back together. With AI, Pega centralized before that could happen.
Where AI spend gets wasted: work a workflow already does cheaply
Asked how small AI-driven improvements connect to the P&L, Stillwell points to cost. He says the industry has an insatiable desire to use AI everywhere, and the risk is using it in the wrong places: where it gives a bad answer, where it does not improve the process, or where it costs far more than a routine process would. His example is replying to an email. The reply workflow is already efficient, and the recipient’s name is in front of you, so asking AI to work it out means paying for it to reprocess something already known. He says this is probably the biggest cost that gets hit.
How to keep pilots from multiplying: think like a product manager
Stillwell compares deploying AI to designing a product. A good product manager building expense management software decides what is core and what is merely nice to have. People who do not think that way will build something impressive with AI that does not help the business and may add risk if it cannot be controlled or accidentally shares information. His advice is to start from the problems a team has, ask whether AI can help, and then design the pilot to prove that out.
What to measure when a full ROI calculation is not worth doing
Stillwell says ROI on AI has been elusive in many cases, and that there is a step before a hard calculation that works as a proxy. If AI could process transactions faster and take two days out of the quarter close, he does not think the ROI needs to be calculated to know there is benefit. If collection notes went out faster and cash arrived five to seven days sooner, a back-of-the-envelope estimate on days sales outstanding would show the value for the year. The calculation matters less to him than thinking through the problem and the outcome first.
Why shutting down an AI project should not count as failure
Stillwell says the emotion needs to come out of AI decisions. A board, CEO, or manager should not be asking for an AI use case so they have one to talk about. The better question is whether the team has thought about how AI will help the business. He gives credit to people who can say they thought AI would help somewhere, found it would not, and moved resources to another project. Treating every use case that did not work as a failure, he says, makes people unwilling to admit when an initiative is not delivering.
When will AI stop needing its own ROI case?
Stillwell expects parts of AI to become natively embedded and obvious, the way nobody now weighs email against driving ten miles to tell a vendor something. He thinks that is a few cycles away, because it is not yet instinctive where AI obviously belongs and the risk profile is not fully understood. Email has had roughly a decade of security hardening. He says AI has probably not had a year.
FAQ
What is the biggest cost in enterprise AI spending, according to Ken Stillwell? Using AI for routine, structured, rules-based or workflow-based tasks. Stillwell says a workflow, once set up, can run billions of transactions very cheaply, so paying for AI to do the same work is where cost gets hit hardest.
How does Pega handle internal requests for AI? Pega channels them through a small AI steering committee that represents all functions, anchored by its engineering AI lead and its CIO. Stillwell says this was done from the beginning to avoid teams procuring their own tools, as happened with cloud.
How can a finance leader estimate AI ROI without a full calculation? Stillwell suggests an intermediate outcome as a proxy, such as taking two days out of the quarter close or collecting cash five to seven days faster. A quick back-of-the-envelope estimate on that outcome gives enough direction to decide whether to start.
When should a company shut down an AI project? When the team concludes AI is not going to help in that use case. Stillwell says those resources should move to a project more likely to pay off, and that boards and executives should reward the people who make that call.
Will AI eventually stop needing a separate ROI case? Stillwell thinks so, once security and governance practices mature and it becomes obvious where AI belongs. He says that is a few cycles away, because AI has probably had less than a year of the kind of hardening email has had for a decade.
What does Ken Stillwell expect the AI conversation to be about in a year? How to choose between a commoditized AI model and an expensive, purpose-built frontier model. He expects two classes of AI to emerge, with model providers differentiating on specific use cases..
About Ken Stillwell
As Chief Operating Officer and Chief Financial Officer, Ken Stillwell is responsible for finance, investor relations, legal, facilities, information technology, Pega Cloud, global client support, and business operations, overseeing more than one thousand employees globally. His leadership has been central to Pega’s transformation to an as-a-Service business through Pega’s Cloud transition. Recognized as incisive, transparent, clear, and empathetic, his opinions and counsel are sought after by peers, clients, and media. Ken is a regular contributor to numerous industry publications regarding financial and operational leadership.
Ken has nearly 30 years of financial leadership experience in high-growth organizations in the technology sector. He is a graduate of the University of Pittsburgh, where he earned his undergraduate degree in business and economics, and the University of South Carolina, where he earned a Master’s in accounting and finance. Ken is also a certified public accountant.
———- Resources ———-
Pega delivers the platform to reimagine, run, and evolve the processes and decisions an enterprise can’t afford to get wrong. We combine AI with proven architecture to keep mission-critical operations governed, scalable, and continuously adaptable. Since 1983, the world’s largest organizations have trusted Pega to turn transformation ambition into durable results.
The hardest part of martech right now is telling which tools are actually earning their keep. That’s the question running through the MarTech Summit Boston agenda on October 27. We’re attending MarTech Summit Boston on October 27 as a media partner. If you want to join, the code AGILE15 saves you 15%.
We’re proud to be a media partner for #MAICON26 – Oct. 13-15! Learn how AI can power your marketing and business and help you grow smarter. Use code AGILE150 to save!
Need to evaluate martech, AI, or CX platforms and don’t know where to start? The Agile Brand Guide List is now in beta. We’re building the most useful, free-to-use, and up-to-date way to assess what truly matters when evaluating platforms.
Reach your customers with Reddit. Spend $500 in ad spend, get $500 back in ad credit!
Chaser is the only Slack-native project management platform that helps teams turn messages into tracked tasks, automate follow-ups, and maintain team-wide visibility, without adopting another tool. Now integrated with Claude and other GenAI tools. Learn more at trychaser.com and use code AGILEBRAND for a 3-month free trial (normal trial is 14 days).
Start building your own apps with Replit and get $20 off.
Enjoyed the show? Tell us more and give us a rating so others can find the show.
Don’t miss a thing: get the latest episodes, sign up for our newsletter, and check out Martechipedia, the wiki for marketing technology.
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content.
Transcript
[00:00:58] Greg Kihlström: [upbeat music] Hi, I’m Greg Kihlstrom, host of the Agile Brand, and here’s a question for you. There’s a line item in your budget called AI. Do you actually know what’s underneath it? Agility here isn’t about moving faster on AI. It’s about being able to tell quickly and honestly which of these investments is working, and being willing to stop the ones that aren’t. Today, we’re looking at what it takes to put real financial discipline around AI investment, including why AI as a budget category hides very different decisions, costs, and risk profiles, how to trace a specific AI-driven decision back to a line on the P&L, and what it takes to scale a pilot across the enterprise without Greg Kihlström: the company losing a single consistent logic. To help me discuss this topic, I’d like to welcome Ken Stillwell, CFO and COO at Pega. Ken, welcome to the show.
[00:02:55] Ken Stillwell: Hey, Greg, good to be here. Thanks for having me.
[00:02:58] Greg Kihlström: Yeah. Looking forward to talking about this with you. Before we dive in, though, why don’t you give a little background on you and your role at Pega?
[00:03:05] Ken Stillwell: Sure. I’m, the chief operating and chief financial officer at Pega. so what that means is, I typically– I generalize that by saying, if, I don’t own the team that, that builds our product, and I don’t own the team that sells and markets our product, though, a lot of the other pieces of the business reported to me. So, so it’s a, it’s a, I would say it’s a broader role. I find myself kind of, mostly kind of building connection points between the building of our product and the selling and the operating of our product and where there may be, places where I can make sure that we connect it across the value chain.
[00:03:42] Greg Kihlström: Yeah. Got it, got it. And for those listeners that may not be as familiar with Pega, can you share a little bit, what’s the, what’s the company’s core focus and mission and, and who are the typical customers?
[00:03:54] Ken Stillwell: So, in large organizations, quite frankly, organizations of even mid-size organizations, there are processes and parts of the business that need to be repeatable, that need to be predictable, that are, that, that, anchor on certain workflows. An example of a, of a, of a workflow that we all know about is, is an ERP system. Now, Pega doesn’t sell an ERP, but if you think about an ERP system and then think about things like managing a customer complaint, managing an exception, a loan origination in banking, onboarding a participant in healthcare, these are all workflows that clients need to manage, many of which are regulated and have to be done in a predictable manner, and they’re done at scale. So we help our clients build systems that help them automate those, deal with certain regulatory
[00:04:39] Ken Stillwell: and compliance, and then more recently, leveraging AI to try to automate and use AI as much as possible to reduce human time that might need to be spent when managing the, those activities.
[00:04:51] Greg Kihlström: Got it. Got it. Yeah. And that’s, that’s, I think that’s a great segue into what we’re gonna talk about today. And we’re gonna talk about a few things and, but I wa- I wanna start with certainly, we talk a lot about AI on the show and AI adoption, but that is a very broad umbrella term for a lot of things, right? So, uh-
[00:05:12] Greg Kihlström: … request lands on your desk labeled, quote, unquote, “AI,” but it could be a rules engine, a predictive model, a system, an agentic system acting, semi-autonomously or autonomously. That’s a v- those three alone are very different cost, risk, payback profiles. So just to start at the top, how do you get a team to really tell you what they’re actually asking you to buy when they say, “We need AI”?
[00:05:40] Ken Stillwell: So one of the things that we’ve done at Pega is we’ve tried to create some centralization of anything that comes through around AI to, there’s two big players. We have one in our engineering department that leads our AI initiatives, and then our CIO, who actually manages a lot of our vendors and instrumenting that, connecting the data, et cetera. So, if anything does come in as a, kind of a, an AI thing, it, it, typically we try to channel that through, the, this AI steering committee, so to speak, that we have. It’s a very small number of people that are, that represent all the functions. So that’s something we did for the exact reason you mentioned, Greg, because if not, people would be out, trying to procure their own. Very
[00:06:25] Ken Stillwell: similar to the way it happened when the cloud came on, where people would try to get credit cards and get an AWS instance, and then you get to some point where you say, “Well, hold on a second. We can’t have cloud environments everywhere. We have to centralize that.” We kind of started from the very beginning with AI.
[00:06:41] Greg Kihlström: Yeah. Yeah. Well, and, and to talk a little bit about the, the, the workflow, components and some, how some of your customers are thinking about this as well. a lot, a lot of businesses, there’s an AI business case that talk about, a- almost operational change is an afterthought. it’s, it’s a tech purchase-
[00:07:05] Greg Kihlström: … or a product purchase first, and certainly important, but, where does the gap between buying and implementing a technology platform and actually operationalizing it show up first and, what, what’s the tell that the, that a case is, is going to have some challenges?
[00:07:25] Ken Stillwell: Well, so one of the things that y- I think you have to accept, and maybe we’re getting, at Pega we’re probably getting to the other side of this, but in the early days we just really wanted people to use AI. We just wanted people to play with it. We wanted to reduce the barriers to adoption, because there’s a big change management aspect to using something new. And so the, we felt like if we created too much overhead in trying to just use some basic tools and that, we might have slower adoption and slower change management, which I think we all recognize in the industry is gonna be hard enough as it is. So at the beginning, I think we just focused on, let’s, let’s expose people at different tools, let’s give them accessibility, and let’s create maybe some governors on what they can actually
[00:08:11] Ken Stillwell: do with those tools so that, they don’t, the, the cost doesn’t get out of hand. But we fe- we first focused on just exposure. we’re now mo- we’ve moved kind of past that to the phase of really trying to, to be more crisp with what problems are we trying to solve, and, and how are those problems gonna help our business or help our clients or improve our solutions that we sell to our clients? And, once again, a much more complicated analysis than just letting people get exposed, but that’s kind of, we’re at that phase of the journey.
[00:08:45] Greg Kihlström: Yeah. Yeah. Well, and so from a, from, from your customer’s standpoint, I know they’re getting a lot of pressure to adopt personalization, personalized recommendations, all of those kinds of things. I, I, my first exposure to the term next best action was actually through some documentation from Pega, so b- this was many years ago b- at this point, but, I know, I know you guys have been talking about this for, for a while. Next best action improvements and incremental improvements are certainly going to, make a change but, a lot of times, speaking as a, as a CFO maybe, they’re, the fi- a finance team is tracking revenues and margins and kind of some of those downstream
[00:09:30] Greg Kihlström: impacts. How do you connect the dots between, okay, yeah, we’re making these, these small incremental changes, they’re, they’re, they’re improvements, but how do you make, connect the dots between those and really the, those downstream P&L improvements?
[00:09:47] Ken Stillwell: I think this is one of the biggest challenges that the industry is dealing with, which is there’s an, kind of an insatiable desire to use AI everywhere you can.
[00:09:58] Ken Stillwell: And the risk with that is that you will use AI in the wrong places, and when I say wrong places, it could be where you get a bad answer. It could also be where it doesn’t actually make the process better, and, and/or the cost is much higher to actually execute AI than it could be a routine process. So there are… So I think understanding those differences and appreciating that AI is a great tool, a powerful tool that can be used as, as an accelerant, but does not actually fit every use case for, is, is really important. And a perfect example of one is if, if you were going to, send an email
[00:10:43] Ken Stillwell: communication, and you had a, a person sent you an email and you wanted to reply to that person, we have a pretty efficient way to do that, right? I think, Microsoft and G- and Gmail, et cetera-
[00:10:58] Ken Stillwell: … have created a pretty structured workflow around reply, giving you r- But if you used AI, for example, if you took that and said, “Based on this email, I would like for AI to decide who I reply to,” and what would be the con- It would go and process around a reply where you already know who the person is, because the name is right there in front of you, so, so because AI has to rethink and reimagine and reprocess, every one of those. So when you have things that a CPU could do or a v- or a task or a rule or a workflow could do incredibly inexpensive, why would you use AI to do that? You wanna use AI for the things that can’t be done through a workflow, right? And, you know,
[00:11:44] Ken Stillwell: and, and like reimagination, creativity, right? things like, uh-
[00:11:50] Greg Kihlström: Taking con- taking content from a bunch of conversations and putting together a nice two-paragraph summary, that’s not something that workflow would do. That’s something that AI could do. So there are things that you would certainly wanna leverage it, but I think that’s the part that I think is, probably the biggest cost that gets hit, is people using AI to do routine, structured, rules-based or workflow-based tasks, where you, you really set up the workflow and it’s incredibly cheap to do billions of transactions afterwards. Why would you wanna do that with AI? And I think that’s– I think the market is moving to understand that more, more holistically, but I think six months ago or nine months ago, I do think
[00:12:35] Greg Kihlström: there was kind of at least a, a, maybe a, a hope that AI would just, be able to just, could do everything. And I think we’ve realized that it’s powerful, but it’s, and it’s actually not efficient to do lots of things that can be done at scale. Yeah, yeah, definitely. I mean, there’s not, not only is I… I think we went from token maxing to tokenomics pretty quickly over the summer-ish, let’s just say. But, so, there’s not only the cost, but the, there’s anoth- the other cost is, is risk, right? Because when you don’t need to generate, I mean, partic- and we’re talking a lot about highly regulated industries too where, no- language needs to be exact and thi- like, we don’t need to regenerate something, to your point, that can
[00:13:20] Greg Kihlström: be very easily rules-based and, and just, kind of, kind of run, set, set and forget almost. And then, yeah, to- agree. A new landing page turns into a pile of tickets and handoffs. Framer helps your team move faster. Teams can collaborate in real time, iterate on the same page, and publish instantly. Agents help you go beyond the vibe coded site and a first draft to bring you a production-ready site faster than ever. Framer is the AI website builder that helps creators, teams, and businesses ship production-ready sites faster than ever while getting every detail right. Prompt, inspect, edit, and publish in one place at a whole new pace. Agents and Framer work alongside
[00:14:05] Greg Kihlström: teams to streamline collaboration on the same canvas, build custom code components, create and manage CMS content, optimize SEO settings, audit for issues, and ship everything all in one place. Learn how you can get more out of your site from a Framer specialist, or get started building for free today at framer.com/agile for 30% off a Framer Pro annual plan. That’s framer.com/agile. Rules and restrictions may apply.
[00:15:07] Greg Kihlström: The other part here and maybe facet of this is just what I run into a lot as well with the, with the orgs that, that I work with, is there’s definitely a, a focus on, okay, let’s, let’s do a pilot. Let’s, let’s figure some things out. And I, and I think, doing a pilot before doing a, a huge effort, it makes a lot of sense. I think another challenge, though, is there’s a lot of pilots all over the org, and you get this kind of fragmentation, right, of like everybody me- means well. They have, they’re, they’re do- hopefully they’re doing things in the, in the best way possible. But, h- as a, as a COO, how do you, how would you recommend someone think about how do we keep this from getting out… Like, we want people to learn. We wanna grow as an organization, but
[00:15:52] Greg Kihlström: how do you keep that from getting out of control and, and fragmenting the organization? I, I think, if, if you draw the an- if you draw the analogy to being a product manager that might design a car, design a software application, the product manager has to understand the value that gets created by decisions that they make. They have to think about, I’m trying to build a expense management software application, so what is the, what are the things I need and what are the things that I don’t need that, quite frankly, might be nice or cute, but, like, they’re not really core to managing an expense processing in a application like Concur? So if you’re a good product manager, you’re gonna, you’re gonna be thoughtful about the, the right outcome,
[00:16:37] Greg Kihlström: the value ch- things that you build in that product. If you draw that to AI, if you have people that are not thinking like a product manager around the part of their business, they will just do things in AI because they can get some whiz bang thing. They can get something that’s, like, really cool and… But at the end of the day, it might not actually help the business at all. In fact, it might put the business at further risk exposure if it’s actually doing something that maybe might, might not be able to be controlled or might accidentally share things. So I think it’s really important to think like a product manager, right? So if you’re someone, if you’re the manager of a collections team in finance, you should think about what are the problems that the collection team has,
[00:17:23] Greg Kihlström: and focus AI on those problems and how you can scale those problems. If you’re working in sales operations, you should think about the same thing around the selling team. But to build just kind of, kind of, um
[00:17:36] Ken Stillwell: interesting things with AI that don’t really solve a problem and you’re not really sure what it solves, that’s where the pilots get out, get out of hand, right? And I do think there’s a lot of pilots going on that are driven by non-product management type thinkers, people that are more engineering that might just say, “I’m gonna just try this ’cause I wanna prove that it can be done,” not actually knowing whether it would actually make any sense at scale. So I would advise companies to think more around what problems are you trying to solve, and can AI be helpful in that? And if so, what would the pilot look like to prove that out? I think you’ll get a lot more value from your spend and, quite frankly, your management focus. ‘Cause it’s not just the tokens, it’s also all the people that you shift
[00:18:21] Ken Stillwell: on to be able to work on that.
[00:18:23] Greg Kihlström: Yeah. Yeah, I know. It’s, I think we all got a set of new toys a, a couple years ago, and everybody wanted to play with them and, and do all that stuff. But yeah, I, I feel like, t- to your point, I, I think there is a bit more pragmatism occurring. Not, I wouldn’t say everywhere and, and all things, but there, there’s a bit more of that. And, maybe that brings us to a little more of, the, the financial and, and the investment component of this as well. And so, there, there are some real bene- When, when things are done well, there are some real benefits: reduced risk, faster decisions, better judgment under pressure. And yet those are … They can be hard to forecast in a financial projection
[00:19:08] Greg Kihlström: and, and things like that. So how do you, how do you think about things like that, smart initiatives and, and, and things like that, and yet a little hard to measure where, as you’re balancing some of those, those other maybe weaker projects that, can’t really justify some of those things?
[00:19:30] Ken Stillwell: So getting ROI around AI has been elusive in many cases. and so-
[00:19:36] Ken Stillwell: … without being able… And, and sometimes, to be honest with you, sometimes instinctively you know there will be value, and trying to calculate the ROI might not really get you much help. But maybe you’re better off to just get started and to try to prove some things out, because you may be so committed to understand a problem, that your instinct is probably good that there will be some ROI from that. But I think-
[00:20:00] Ken Stillwell: … there’s, I think there are, there are like, there’s a step before a hard calculation of ROI that I think can be a good proxy for ROI. For example, if you knew that AI could help you process transactions faster to be able to increase the close at the end of a quarter by two days, I don’t think you need to go as far as to calculate the ROI on that. You know that two days of a close is gonna have benefits. It’s gonna have benefits-
[00:20:31] Ken Stillwell: … all kinds of efficiencies. Now, mi- maybe it’s not at the level that it’s worth sustainably doing that with AI, but at least you have a, a, like an instinct or a direction to say, “I’m really gonna try to take a couple days out of the close.” Another example might be if I believe that getting collection notes out faster, I’ll receive the cash by five to seven days faster. I don’t know that going through a complicated mathematical ROI calculation. You could, you could do a back of an envelope and say, “If we could reduce day sales outstanding from, 60 days to 52 days, this is about what it would give us in value for the year.” So I think there’s ways that you can either do an i- kind of an intermediate outcome before OI… or ROI,
[00:21:16] Ken Stillwell: or you could just estimate ROI in a really quick way to say, “If we could get this outcome, I, directionally, I think this would be the value.” So I think there’s ways that you could… But the key is to think about what the value would be. The calculation to me is not as important as thinking about the problem you’re trying to solve and the outcome you wanna get, as opposed to just doing AI and hoping that you will discover ROI on the other side. [laughs]
[00:21:45] Greg Kihlström: Right. Right. Yeah, no, u- unfortunately, I think there’s a little too much of that still. But, a- and I think there, there’s another thing too, and, I think a lot of the conversation is about, how do we invest in AI, and what’s the best way to do that? But I think there’s another component of this, and I read a blog post, the other week on just, nobody gets promoted for shutting down an AI project or, being a little snarky there. But how do you determine… again, there’s so much pressure even from boards, from execs to do AI, quote, unquote, and yet sometimes any initiative, it, it’s, in order to do an experiment, 100% of experiments are not gonna be successful, right?
[00:22:30] Greg Kihlström: So how do you-
[00:22:32] Greg Kihlström: … how does a leadership team start thinking about shutting down an AI investment in this, in this kinda climate of we gotta do more?
[00:22:42] Ken Stillwell: I think the… We’re not there yet, but I would say I think you’ve gotta take the emotion out of this, right? You should not be deploying AI. A board should not ask you about AI. A CEO, a manager, a leader should not be saying, “let’s just do AI to do AI.”
[00:23:00] Ken Stillwell: Or, “Hur- hurry up and show me some, AI use case so that I can tell others.” I know that’s the human behavior to be able to, like, we’re competitive, we wanna answer the question, like, are we a think- are we a forward thinker? Are we being innovative? I think the questions need to be, have you thought about how AI is actually gonna help your business? And by the way, you might be wrong, and I give a lot of credit to people that actually can acknowledge, I thought AI was gonna help me here, but it isn’t actually going to be the case. I’m gonna shut that down and shift resources over to another project that I feel like might actually get me further. I think- I think boards and executives should reward team members for having that thought
[00:23:45] Ken Stillwell: process of, like, where’s the real value, and, and not view it as, “Well, you couldn’t make every use case work, so you must, you must have failed.” Because I do think that’s a, that’s a trap which will make people then not want to acknowledge that they’re not getting the value out of, some of their initiatives. And you, you and I just talked about earlier, Greg, at the beginning we’re just really, we’re kind of just figuring this out. So to think that some of these things would have to be shut down, yeah, it just seems illogical to me. Of course we were gonna guess wrong on some things.
[00:24:18] Greg Kihlström: Right, [laughs] of course.
[00:24:19] Ken Stillwell: [laughs]
[00:24:19] Greg Kihlström: Right, right, yeah. Well, and, and I think, another part here is, I don’t think we’re quite there yet, but as AI… I mean, every conversation on this show involves AI in some degree, even on, topics that you wouldn’t necessarily think. But AI, as it becomes, more embedded in core operations, like, do you see a f- do you foresee a point where measuring standalone ROI on AI is just sort of… I, somebody wrote an article a few months back, like, AI is gonna be like electric. we don’t say electric refrigerator anymore [laughs] or anything like that. It’s just a refrigerator. w- again, I don’t think we’re there yet,
[00:25:05] Greg Kihlström: but, do you foresee a, a point where we’re really just talking about the ROI of the thing and, what, what does it take to kind of get to that point?
[00:25:15] Ken Stillwell: Well, I, I, look, e- email is a great example in that camp, right?
[00:25:18] Ken Stillwell: I mean, we don’t try to… We don’t say, “Well, stop, stop, driving 10 miles down the road to a vendor to tell them something when you could actually send them an email.”
[00:25:29] Ken Stillwell: Like, we don’t even think about that anymore.
[00:25:31] Ken Stillwell: So I, I think, yes, of course, there will be aspects of AI that are just natively embedded, and it will be obviously you do that, right?
[00:25:39] Ken Stillwell: Obviously you use tools like Microsoft Word and Microsoft Excel and PowerPoint because they’re, they’re better than creating them from scratch. I mean, so I, I, I don’t think there’s any doubt in my mind we will, we will move in that direction. I think that we are probably a few cycles away from that because we don’t really instinctively know where AI is just like, obviously you should use it there.
[00:26:06] Ken Stillwell: And we don’t really know about the risk profile completely. And I think understanding the risk profile and hardening that to make sure that when you send an email it only goes to the person you sent it to, when you actually have an email on your computer everybody else can’t access that email. Like, there are things that we’re comfortable with because of, quite frankly, a, a decade of security and hardening around that, and we haven’t had that decade around AI.
[00:26:34] Ken Stillwell: We probably haven’t even had a year of hardening around AI. So I do think, when we talk about AI security and governance in, in the, in the political discussion right now, I, I don’t mean that. What I really just mean is just good best practices around how AI should be governed and managed.
[00:26:53] Ken Stillwell: Because you cannot just, allow it to do whatever you want. You, it has to be, with a, a purpose in mind and with a security. So I think once we get the security standards in place, I do think we’ll, we’ll get comfortable that AI is just obviously used in these use cases. And I think we’ll probably not be as, kind of obsessed with the ROI calculation ’cause we, we will just know that it has helped our lives.
[00:27:21] Greg Kihlström: Yeah, yeah. Got it. Well, Ken, thanks so much for joining today. I’ve got two last questions as we wrap up here. The first one, if we were having this interview one year from today, what is one thing that we would definitely be talking about?
[00:27:34] Ken Stillwell: I think we would be talking about, how to, how to pick when you use a commoditized AI model and when you use a frontier, very expensive, purpose-built model. I think that there’s gonna be, I think we’re gonna really lock down on i- it’s electric, and whether you get electric from one company or another, it’s the same.
[00:27:56] Ken Stillwell: But when you actually need a more powerful model, that’s where I think the model providers will differentiate themselves, around specific use cases where they’ll be better than anyone else, ’cause I think there will be two kinda classes of AI. We’re not quite… We’re moving in that direction, but I still think it’s gonna take some time for that to play out.
[00:28:13] Greg Kihlström: Yeah. Got it. And last question for you, what do you do to stay agile in your role, and how do you find a way to do it consistently?
[00:28:21] Ken Stillwell: I have found that one of the most important resources that I have are other people, and actually talking to them and what are they doing in their businesses and what are they hearing and what are… And so I try my best to talk to as many people as I can outside of Pega and just learn from that and w- look for the patterns and try to do pattern recognition on what I’m seeing and, and then educate myself from that. and I, I found that human experiences are really good learning, and I, I only have enough in the, in the, 16 hours or so that I’m awake every day, but if I talk to 50 people, I’ve multiplied that by a lot. So I think that’s valuable.
