Responsible AI isn't an optional layer, it must be foundational - Interviews from Pegaworld 2026 Pt2

0:00Welcome to the next edition of the PunkCX podcast. This podcast is part two of the two-parter that features a series of chats that I had with Pega executives and one of their clients while at Pega World in Las Vegas recently. In this episode, I talk with Matt Healy, Senior Director, Product Strategy and Marketing at Pega and Tara DeZao, Senior Product Marketing Director at Pega as well. Now, some of the things we cover include the big themes and takeaways from the event, the future of customer service, ethical AI, the customer engagement blueprint, and the new and exciting agentic capabilities of customer engagement studio.

0:33Let's get into the first conversation that I had with Matt. Welcome to this next installment of the PunkCX podcast. We're here at Pegaworld. This is going to be installment number two. You're going to get paired with an interview and a chat I had with Tara, which is going to follow a podcast that I did with Ken and one of your customers, Peter Lacroix of Achmier. He thinks he's great. Anyway, so welcome. Welcome back to the podcast here. For the people who don't know you, obviously, Matt Keely, Senior Director of Product Strategy and Marketing at PEGA. Tell me, what does that mean? Yeah, yeah.

1:10Thanks for having me. Super excited to be back. Honored to be in a one-two punch with Tara. That's awesome. Good pairing. So yeah, I lead a go-to-market strategy for the Pega platform, which means I get to think about how we are helping partner with enterprises to deploy AI into the ways that they build applications, so into their development, and then into the way that they run their business. So deploying AI confidently into their operations, their service, and their customer engagement. Awesome. Now, we're here at Pega World. We are now winding down end of day two. Things are starting to percolate.

1:46ruminate sort of there's lots of been lots been going on and what have been your main sort of standout highlights of the the last two days yeah i can't believe it's almost over crazy two days jam-packed but you're right we are coming to the end everyone's hitting the bar hitting the chess tournament uh which we which we wrap up with here so yeah tons of announcements i would say like what has stuck out to me is the theme that we've had this whole event predicated around is the idea of helping enterprises deliver predictable outcomes at predictable costs, unlocking predictable ROI with AI. So that's using AI, of course, to accelerate development, of course, to automate work, but helping enterprises deliver on the considerations which are unique to them.

2:31So the idea of making sure their rules are hit 100 % of the time, making sure their regulations are hit 100 % of the time, making sure you know they can engage their customers in accordance to their brand with trust with empathy yeah um so you know that's sort of been the theme and tons of amazing product announcements and go-to-market announcements all sort of laddering up to that awesome and i know that they've kept you busy because you've been speaking in a number of different sessions let me get this right you've been talking about navigating the future of customer service a strategic roadmap that sounds exciting and then the one which is like i hope probably the the has the best headline is like vibe to live 8x faster with go live with with pega so can you give me the um quick summary on what those are the big themes are and one the future of customer service and that roadmap and also vibe to live yeah absolutely and it touches on the the two um you know sort of high level themes that i talked about so the future of work customer service and operations um we introduced some new capabilities that we this week that are all about extending our workflow automation capabilities, the ability to drive end-to-end outcomes, automate an end-to-end customer journey.

3:41That's now extensible to be plugged into any agent. So what that allows enterprises to do is to take their North Star vision, which is right now engaging clients, engaging customers, engaging their employees through agentic experiences, Right. Think, you know, calling into agents to get serviced, engaging through ChatGPT to, you know, pick up on insurance or, you know, a file for a loan. They're now able to take their products, their services, their processes and plug them into those agentic experiences in a way that is going to deliver on, you know, their compliance and deliver 100 percent consistency 100 percent of the time.

4:21Right. And is that, you're doing that, is that through the MCP server sort of integrations or the facilities, the functionality that you've developed? That's what you mean. Yes. So now every process, every workflow you build in Pega is exposed to any agent through MCP. Okay. And so, I mean, what does the future customer service actually look like? I mean, because there's a lot of talk about it and, you know, with this influx of AI and kind of automation and changing job roles and all these different sort of things. People talking about self-service, automating all the simple stuff and then lying kind of agents to focus on human agents to focus on the more complex stuff.

4:58I mean, but that's kind of where the current state is, right? You're talking about thinking about the future a bit further out sort of thing. What do you think? What do you see as being the future? Yeah, I think it's sort of two things. in my mind. So one is agents will be the front, AI agents will be the front door to your enterprise. Right. So that's, you know, and we've seen, already started to see customer behavior shift to procure through, you know, AI agents. Like, for example, if I want to file for insurance or if I want to, you know, get a loan, I'm not starting on a website anymore. I'm not starting in Google.

5:39I'm starting in ChatGPT, right? And I'm asking, what are the right products and services for me? And then, you know, we're starting to see more enterprises plug into those interfaces to actually not just surface their products, but, you know, walk that customer through being onboarded. And eventually, you know, that'll extend into service. So, you know, that front door aspect is where we're going. But agents right now, AI right now introduces variability, introduces hallucination obviously there's this whole idea of token economics which we can get into yeah so enterprises don't have a an approach that allows them to plug their services in that is going to enable 100 predictability in the ways that they're actually interacting with their customers right so that's what we're focused on so i mean i think it's interesting you mentioned the tokenomics and um i think it's a fascinating thing because it's like it's always struck struck me is that that the economics of this are still not clear.

6:40They're only just becoming clear. We've sort of seen people getting kind of big bills and they're going, there's a shock sticker price sort of thing. But it also seems that everybody's rushing into sort of more generative AI sort of approach. And the way that even when the current economics become clearer, there's not a lot of economies of scale in terms of your investment and taking that sort of approach. And so do you think that we're going to have to go through a maturation about how people approach kind of things? Or are we just going to have to almost fine-tune some of the tools such that we're not always running the same query a thousand different times and possibly getting a little bit of variability, but paying for a thousand different times rather than just going, oh, that's the right one.

7:26Why don't we stick on that one? Do you know what I mean? Yeah. Yeah. Obviously, you know, like anything, it comes down to using the right tool for the right job. You're right. And, you know, what we've seen to this point is enterprises are obviously in a race to adopt AI. They see opportunity. So they've invested tons. They've started to build out use cases. But nobody yet is actually realizing a positive ROI. And one of the reasons why is because agentic systems, like if I'm going to build out, you know, a customer service workflow, I want to be able to onboard customers using AI. Right now, I'm probably going to build it using an agent to orchestrate that process.

8:01Okay. Right. And then every time a new customer comes in, that agent has to think through, reason, churn through tokens, how it's going to service that customer. And then as it actually processes that work and gets work done and calls these sub-agents, what happens is it re-reasons every time. So it's using the entire context of its work and churning through tokens as it walks through that process, which means that, you know, for simple workflows, maybe, you know, it's not going to cost that much. And I have a positive ROI. But as I go through multi-step workflows, complex work, the economics make it so, you know, it's going to cost me multiple dollars to process every customer interaction, which obviously is going to put me in the red for most use cases.

8:46Yeah, yeah. And compound that with the fact that AI vendors are, you know, the pricing that they have is right now completely unpredictable. So it started off, you know, last year, multiple years ago, where tokens were extremely cheap. And these AI vendors were sort of subsidizing it. They were giving it away. Still are. Still are to some degree, right? But we've even seen recently a lot of vendors who are doing more cheaper token prices or even all you can eat models increase their price or shift their pricing model. So where a use case I might have built before was in the black, today it might be in the red, which is completely unsustainable for an enterprise.

9:26So our approach is all around licensing, not by token, not by how much thinking the AI does, but by how much work you're actually getting done. Yeah. So the end value. And that's our sort of case-based approach. And but that's also going to be also disaggregating the process because you understand that you have control over different parts of the kind of process. And then you can also then pick particularly, well, the most appropriate tool for the particular part of the kind of process. Yeah. Yeah. So it all comes down to not using AI to reason through and orchestrate the end-to-end process. You know, of course, we're going to use AI within the process to handle documents, to summarize, to research, to automate.

10:07And we love that. And that's good stuff. But using more deterministic logic to actually drive the end-to-end process delivers better token efficiency, which allows us to give all-you-can-eat AI to our clients, but also delivers certainty of outcome, to your point. So it's sort of predictable outcomes, predictable ROI all in one. Awesome. And so that's on that. Vibe to Life. Tell me about that. Yeah, Vibe to Live. So I'm super excited about this one. I came up in the product side of the house. Okay. I was actually like a release manager. I was helping, you know, accelerate our developer productivity within Pega before I moved to go to market.

10:45So this hits right at home for me. Okay. But the fact of the matter is over the past, you know, especially six months, but really like 18 months now, the development world has changed. So 95 % of developers are using AI on a day-to-day basis to move faster. Cloud code, codex, copilot, whatever it may be, which is awesome. And we knew that we needed to incorporate that into our platform to allow developers to move at the same pace while building workflows, building Pega applications. And we think there's a lot of benefits to actually infusing that into our platform. Because what we've seen over the past couple of months is enterprises have sort of rushed to enforce their developers to use AI to build fast.

11:32And similar to the sort of orchestration conversation, they're not getting the value out of it. That's where the token maxing sort of stuff kind of like originated from. or I don't remember which company it was that was talking about they had like a leaderboard about who was going to consume the most kind of tokens and then the bill comes to you and you're like going, oh, maybe not such a good idea. Yeah, I mean, using all your tokens was a little bit of a flex a little bit ago. It was like a cool thing to say on LinkedIn and now like CFOs are getting involved and they're like, that's actually not very cool.

12:06Not when I'm having to pay for it. Yeah, so there's the two sides of it on the development conversation too. There's the price, the token consumption. And, you know, to build an application, you might spend as much on token costs as you would a highly skilled developer in terms of their yearly salary. But then there's also, again, the certainty of outcome idea. Where if I have AI churning away developing code from the ground up, how am I going to be able to audit what the application is doing, make sure it's going to be able to scale efficiently, make sure it's going to deliver on my compliance, SOC 2, GDPR, data handling, whatever it may be.

12:43So what we're allowing enterprises to do is to use any of the leading coding agents, but put them to work not to build code, but to define business logic, define workflows in a repeatable way, in a transparent way, in a way that they can instantly scale. Okay. And that's how you control the costs of it all. Yeah. So there's a couple like, so we are allowing enterprises to build with AI at no token cost in the platform. And, you know, there's a couple reasons we can do that, but essentially it comes down to the idea that our platform is fully model driven, which means you're not building from the ground up using code, but you're putting AI to work to define business logic, like configuration, which of course allows it to be more token efficient, but also more consistent in the way that it builds things.

13:31I mean, and, you know, given that your heartland, as it were, is in regulated industries. And so that sort of level of control and tying things down to kind of process and workflows that you can audit and that you can control is because it's all about certainty. I mean, there was, I think Don was talking in a briefing session and he was like, we have clients that go, you can say to them, oh, you get like 85 % sort of like certainty sort of thing. And he goes like, yeah, but they'll turn around and go, that's a 15 % failure rate. Yeah. And the regulated industry is like, that's not cool. Yeah. I mean, you know, I think one of the contributors to the fact that enterprises haven't seen ROI, besides the whole token conversation, is just the fact that, you know, the areas with the most opportunity for AI transformation are, you know, where are they going to be?

14:20It's going to be manual work. It's going to be semi-repetitive work. It's going to be high-volume work. Where are those areas within an organization? It's customer-facing work. It's regulated work. It's mission-critical work. There's a complete overlap with the areas with the most opportunity being the areas with the most risk if things go wrong. So these are essentially areas where enterprises need to make sure that AI works in accordance to their rules, regs, and standards 100 % of the time. And are you seeing on the service side of things is that because it's the technology we use to automate a lot of the simple stuff, right?

14:58And that's having an impact on the makeup of, well, the sort of the work makeup of service to support workers, right? And are you seeing kind of people kind of like thinking about some, one of the other things I'm thinking about is almost the future of work and the redesign of kind of work, particularly in that kind of context. It's like if all the simple stuff gets automated and we want people to do more of these sort of things. But then the nature of enterprise software has changed because actually the responsibility for management and quality and things becomes a shared responsibility now between vendor and client.

15:35So there's a different makeup of skills around that sort of space. I mean, are you seeing people kind of lean into that or just are they sort of figuring that sort of stuff out right now? because it feels like it's a really emerging sort of like thing like oh my god this is different yeah yeah and uh you know this is above my pay grade if i'm making any predictions on this you're like i'm just they're out of i'm pulling them out of the air essentially right where where is this all going i think we've seen enterprises try and shift their workforce even before ai right um especially in like the service space the operation space everyone's wanted to move towards more generalists rather than specialists everyone's wanted to decrease onboarding time for employees.

16:15AI unlocks some new opportunities there, which we've built in some capabilities and helped enterprises capitalize on. But I think the imperatives have been the same. I think the one really interesting idea that I've started to see emerge is the idea of not incorporating AI to the scale where your enterprise becomes an AI wasteland in the way that you show up to customers. So I think we've started to see enterprises deploy more intelligent approaches, predict how and where they should not rely on AI to interact with their customers, but actually bring a human. And customer service is a great scenario.

16:54Right. Like if someone is having like troubles with filing a claim and they just had something happen to their house or, you know, they just got sent to debt collection and they need to figure out a way forward. Like these are some of the most important times in your customers lives. Yeah. Send an AI agent who's going to first of all, everyone's annoyed with to begin with, but then it gets something wrong. Right. So predicting, you know, those moments, predicting, you know, how you can sort of maintain or even enhance your customer relationships and, you know, show up to your customers is, I think, going to be an emerging trend.

17:28I mean something I've advocated for for a long time is that the end you know you should always have a granular vision of what you want your customer to experience across the piece across different types of different types of customers and also what the human and tech balance should be like in that now it doesn't need to be kind of like crystal clear it just needs to be directionally kind of accurate but if you kind of do that then you take a more i think you have a more sophisticated grasp of um what you need from a human being and what you need from a piece of technology rather than going i'm going to take this and i'm going to figure some stuff out or don't automate some stuff and then get what i get sort of thing and so i if it's an emerging trend then i applaud that because i'd like to see kind of more of that people leaning into that sort of i'm making an emerging well thank you let's kind of keep talking about it and then we'll just i'll say rather than it being a desire, I'll just say it's an emerging trend.

18:26Matt Haley said it was an emerging trend. It must be an emerging trend. That's fine. And then hopefully it makes more people get convinced and then we see more of it happening. Yeah, there we go. Brilliant. So just thinking about a couple of final things, what's coming up that you're really excited about? Oh man, so much to be honest. I think one thing we've started to see pick up that we are continuing to enhance on is this idea of AI-driven legacy transformation. Okay. So we saw, we had AWS and Unum speak on our main stage about how they've started to, so we partnered with AWS essentially to tie in their mainframe COBOL analysis tooling with a capability we have called Blueprint, which allows you to sort of forward engineer solutions.

19:15And Unum, an insurer, they picked up on that and they took their disability claims, which had been running on, you know, 20, 30 year old COBOL. And they were able to get it out of COBOL into the cloud in just three months. And before that, they had gotten a quote for seven years and$25 million to get the same transformation done. Okay. So we've seen a dramatic acceleration there. I think that's going to continue to accelerate where like this time next year, we might be talking about getting off of, you know, getting more clothes out of the mainframe into the cloud in three hours. I don't think is I think it's a little crazy but I don't think it's like out of the realm of possibility the way that things are going so I'm super excited about stuff like that and you know I'm going to try and stay one step ahead of the market and probably fail doing it but you know I'll do my best so one final thing more of a best advice sort of question so people listening to this if they want to improve their customer service or their customer service experience let's say let's keep making it specific what would be Matt Healy's best advice?

20:19You said Matt Healy says do this. I mean you talk about yourself in the third person if you really want to but what would be your best advice? Yeah, I think I'll flip it and talk about something else a little bit rather than prescribing. I think enterprise leaders, service leaders, they know their customers. They know what they need to deliver. I think they're struggling more with how to get there and what AI can do, what it can't do. So I think it's really important for all of us right now, you know, to be realistic about AI. And the best way to do that is to get hands on with it, experiment, you know, take some of these capabilities, try things out, have our teams do the same, have a little bit of a growth mindset.

21:02A lot of bit of experimentation and incubation before trying to do things. I think we're making big bets on AI. We'll see if they pan out, but we got to road test them before we actually push some things out there. So awesome. Yeah. Perfect. Thanks for having me. Thanks very much. Appreciate it. Now, I really enjoyed my chat with Matt. It's always great to catch up with him. And I particularly liked his view on how we should prioritize predictability over generative hype. And also how we should be using AI, not just for front end engagement, but for back end transformation. So let's get on to my next conversation.

21:36And this time I talk with Tara DeZaun, Senior Product Marketing Director at PAYA, where we talk about ethical AI, the customer engagement blueprint, and the new and exciting agentic capabilities of customer engagement studio. Welcome to the next edition of the Podcast Podcast. We are here at Pega World. This is probably going to be installment two of the bunch of podcasts that I'm going to put out. But I'm here with Tara DeZao, Senior Director of Product Marketing at Pega. Yes, that's me. Welcome. Thank you for having me. Again? Again. It's a pleasure every time. Absolutely. So for people that aren't serial sort of listeners, can you introduce yourself?

22:16Yes. So I lead product marketing for Pega's Customer Decision Hub. I've been at Pega for about five years and I've seen a lot of really awesome innovation during that time. And I advise marketers and customer engagement practitioners in various ways. I'm a marketing nerd and I love my job awesome love it um so we're at pegaworld there's been a whole bunch of things going on um i've always i always like to start with asking people what are being the highlights or standout moments kind of for you yeah so there's two really defined standout moments for me the first was rob walker and don sherman's co-presentation today about customer engagement studio both rock stars both rock stars very cool metaphors we'll come back to yeah we'll come back to that and just really exciting to see customer engagement studio on the big stage and how it's going to help marketers and customer engagement practitioners the second one was a real tearjerker it was a client story from nhs 24 in scotland it was the nurses line yeah and how technology is enabling them to be able to serve them.

23:33We must talk about the subtitles on the part of the video as well. Yeah, yeah, for sure. Because it was largely American or an international audience. Yes. But they were speaking in sometimes broad accents and there were subtitles to help the audience kind of like understand them. We have a particular trouble over here understanding Scots. I don't know why. it's just it's just a confusing uh dialect for us for some reason so i think they were trying to help the audience yeah i think that's fine but this story the story itself was brilliant oh it was brilliant i mean and it was brilliant in that it also never mentioned ai once it was more about the people of who it helped and what the systems was allowing them to do and where they were moving from to kind of what they're where they've been able to move to yeah and that was a that was a cool thing so it was more about actually it's about outcomes yes you know i mean another point right that they were making was the person on the phone doesn't know how it works or what's going on behind the scenes and the agent that they're talking to or the nurse that they're talking to doesn't know what's happening it just works so that they can do their jobs.

24:50Yeah, it's there to help people. Yeah, it was great. Awesome. I love that. So also, I want to come back to the metaphor sort of thing and the Donna Robb sort of show, and they had their fancy dress sort of thing going on. We'll come back to that in a minute. But you live, Peg has also had you, kept you busy. Yes. And you've been doing a number of different sessions. Yes. One was on ethical AI and one was like transforming marketing with customer engagement blueprint. And I know there's been some developments in there, but I wanted to see if you can give me the human GPT sort of like summary version of each session, if that's all right.

25:27So the ethical AI panel was about how brands, especially brands in highly regulated industries like healthcare, banking, insurance, are implementing ethics and responsible AI and what that strategy is, how you should think about that. as we had one gentleman from Bupa say, responsible AI and AI are not different because it would be foundational to any AI deployment you have. And how are you codifying that into sort of practice? Or how are they codifying that into practice? Yeah, so there are a couple of ways. So Human in the Loop, Wells Fargo was talking about they have a personalization engine and they treat, every day, they treat like a dipstick in your car and they test the models to make sure that everything's you know on point they think of regulation as guardrails for them okay that orients their strategy and Pavan also from Bupa brought up that you know creating consumer trust is like a staircase because it takes a long time step by step to get to the top where they trust you but then if you violate their trust it's like going down in an elevator or you just or you just fall down the stairs and you hurt yourself yes so you know at Pega we say your AI responsible AI it will be transparent and explainable to a human audience so you can explain every decision it will be fair and unbiased to all groups it It will have empathy.

27:09So, you know, I always say just because you can do something doesn't mean you should, right? Like you shouldn't over-serve somebody. For example, you know, gaming and gambling opportunities. That's, you know, an example that I use frequently. And robust because sometimes AI is not exposed to every situation that it should be. So once it's trained and it doesn't, and it experiences a new situation, if it's not trained on that, it will react in an unpredictable way. But also it doesn't mean to say that you're always fixed. I mean, like Wells Fargo was talking about on the main stage and they were like going, we set it up to go so far, but sometimes the model might suggest like, oh, there's this thing here that might be an interesting kind of thing.

28:01And that's where you end up with the model sort of pushing that. like here's the audience but if you did this then it could be slightly kind of more you have to get somebody to approve that as a one-off sort of thing for sure always never uh you know rob mentioned this too that you yesterday in the media roundtable that you know you don't want to have agents like deploying stuff right into production into the customer without checking no no no you I mean, yeah, that's kind of problematic and also sort of different levels. But then tell me about the sort of transforming marketing of customer and the customer engagement blueprint.

28:42Because let's face it, the blueprint, which is a process and workflow oriented kind of one, seems to get a lot of the spotlights. Yes, for sure. And then it's been out for two and a quarter years. Yes. Well, two years officially, but two and a quarter with a soft launch. And then you've got the customer engagement blueprint, which is about 18 months kind of old. Yeah, 18, about 18 months. And it's, I think, just as transformative. It's very transformative, yeah. But it doesn't get as much of the spotlight as the big blueprint. For sure. So tell me about kind of that and what's been happening in that sort of space.

29:21I'm assuming that's what you talked about in your other session. Yeah, so we had brands really just talking about how they're using it to collaborate better across functional areas, explain customer journeys and decisioning and personalization to folks who maybe don't understand it or haven't heard of it, to ideate and reduce time to market. because you can essentially create visuals of multiple journeys across various channels and various tones that are completely aligned to your brand. So you can actually then take the output of that and either put it into production if it's up to your standards or take it to your agency or your internal agency and say, hey, this is exactly what we want to create.

30:07So it reduces speed to market. And it helps you actually scale your action library because you can create as many journeys in as many tones across as many areas. That's good to what action library is. That's a CDH. That is, I know. I'm guilty of using Pegaspeak and I'm always knowing people on that. It's basically a library of conversations that you want to have with your customer. So, you know, if I only have 10 things to say to you, that's only going to take you so far with your personalization, right? Because that's not very many. If I have a thousand things I can say to you, like dynamic, creative, then I'm going to be able to have a highly personalized conversation with you.