Reggie Townsend: Making Responsible AI Irresistible

A few weeks ago, Reggie Townsend stood on the SaaS Innovate stage and made a room full of executives a promise that being responsible could be the easy choice, not the hard one. He called it making responsible AI irresistible. Today, the promise meets the deadline. This is Scouting for Growth. So welcome back to Scouting for Growth. I'm Sabine van der Linden. My guest today is one of the clearest voices in responsible AI. Reggie Townsend leads the data ethics practice at SAS. In the past, he has advised the White House through the National AI Advisory Committee. Today, he sits on the board of Ecole AI, and he is the architect of a simple, stubborn idea that good governance should not feel like friction. When we last spoke, following a SaaS Innovate keynote, AI governance was still mostly a conversation about principles. A few weeks on, it is a conversation about deadlines. In December, the EU AI Act becomes fully enforceable, the first binding AI regime anywhere. Note that the date was postponed from August 2nd. the product Regis Teams is building to meet that moment. SaaS AI Navigator is still being designed as we speak. And the agents everyone was theorizing about, that is already inside the enterprise, acting faster than the rules can keep up. So this isn't a victory lap. It is a reality check. Regi, welcome to the Scouting for Growth podcast. Hi, Reggie. Thank you very much for joining me today. Thank you for having me. I'm happy to be here. This is a lovely space. It is, right? Yeah, I'm digging it. And, you know, they were built probably a year ago. So when I got to know that I could do podcasting here, I was first one on the list. All right. All right. Well, I'm looking forward to this. You know, when we had a chance to talk briefly when you were at SaaS Innovate, I told you the kinds of things that I typically talk about are very difficult to communicate in, you know, two-minute clips. Absolutely. These are nuanced conversations. And so the opportunity to sit down and have something a little more long form with you is definitely what I was looking forward to. Well, I can't wait. And, you know, when we actually met at SAS Innovate 2026, only just a few weeks from now, I can't believe how fast time goes when you think about it. We talk about responsible innovation, AI accountability, and the intersection of regulation with all the transformation we are going on, which is going on right now. You know, what do those words mean for you? You know, when you think about the innovation we're going through, the regulation that needs to be taken care of right now by a lot of organizations, the governance they need to implement, and, you know, driving accountability within their organization. What are those keywords mean for you? I think dynamism comes to mind for me. You know, there's a lot that's going on, much of what you articulated, plus more. And we are all in a very dynamic, fast-moving space trying to figure out some foundational topics, right? And so as the foundation of times past, systems past, processes past are adjusting, you know, we all have to just accept the sheer dynamism. I think we also have to resist certainty at this point in time. It's not to suggest that we shouldn't pursue something that's more stable and certain, but only to embrace the moment that we're in. And to be cognizant of what it means to be in a dynamic fluid space at this moment in time. You know, this is what I think emergence feels like. And at some point after you emerge, you stabilize, right? You reach a point of homeostasis. But we're not there yet. And I think resisting the urge to be absolute, resisting the urge to be certain right now is really important as we consider the foundational adjustments that need to be made in order to make, you know, tomorrow's society more reflective of the kind of society that we all would prefer to live in. When I speak with executives today, I often tell them that, you know, we used to use the word digital transformation. I think the word, you know, we've been using it since 2011. Today, we are going through a frontier transformation. So they need to cross a frontier to become different organization, enabling their people to be comfortable with the new technology which are brought to them, but also build an environment which is a little bit like the second industrial revolution where electricity was put underneath the ground. Right. We are doing this now with AI. Right. We are going through an AI revolution and we need to actually create the pipe to meet to make that AI possible. In the right way. So when we think about responsible AI in 2026, what does it mean for you, Reggie? Well, for me, it means that we no longer should be talking about if. We no longer should be talking about whether this is a necessity. Instead, we should be talking about how. We should be questioning for what purpose, to what end, and for whom might AI fail. And I think we should be orienting ourselves around this idea of using these new sets of capabilities that are being created on the frontier, as you describe it, in ways that reduce suffering for humans, right, first. I think, you know, that would be the principal orientation that I would suggest responsible AI invokes for me. Yeah, we're going to make plenty of money. We're going to do some really cool things, all of that. But if during the process we end up marginalizing, if in the process we end up harming, then we have to ask ourselves if it was worth it. And exploring those trade-offs before implementation, I think, is at the root of this notion of responsible and trustworthy AI. So first, thank you for inviting me at SAS Innovate 2020. It was my first SAS Innovate, and I was actually really impressed. I will use the word impressed as to how you care for your attendees, your customers, the amazing people I've met there, actually. also going to dinner with some of your clients. So thank you for the gift. And, you know, 3,000 people attending Sassino Day 2026. And when you went on stage, you highlighted a few things which stays with me. You said governance has to be the path of least resistance. You want to make being responsible irresistible. And we even had a little bit of a clip for social media, which, you know, has gone crazy. So when we look at this, now we are here together, diving into that statement a little bit deeper. Has Irrestrible survived contact following when we met a few weeks ago? I think so. I think so. Now, I'll admit, you know, I'm a poet, okay? So I'm taking a little poetic license here. And so making being responsible irresistible is as much of an aspiration as it is really an instruction to my team. And, you know, when we sat down to conceive what, you know, eventually became SAS AI Navigator, which I assume you'll want to talk about at some point. What I wanted them to pursue was not just the process of creating a technology, if that makes sense. I did not want them to just create yet another tool for people to have to adapt to. Rather, we wanted to approach this idea of lowering the barriers for responsibility. Like oftentimes the idea of governance, or we'll call it responsibility, comes at the tail end of something that has been created. And understandably, it then becomes a roadblock. It becomes a necessary next thing to do that no one really wants to do. But I said, what if we could turn that on its head? And what if we could inspire people such that they want to be a part of being responsible? Don't even call it responsible. Just call it part of the work to be done. And so what that idea did was sparked design questions that were different. What it did was sparked questions about whether we actually needed a certain level of detail that we thought we needed. And it just kind of forced us to think differently about this notion of effectively a governance product. Right. And everything from the aesthetic. Right. I tell them it's got to be beautiful. People are drawn to beauty. It's got to be convenient. It's got to be intuitive. And I think they did a really good job at accomplishing those goals. Now, again, at the end of the day, we're still talking about technology. We're still talking about the need to check a few boxes. But if we can lower some of the barriers that exist, I don't think that people are, you know, intentionally not being responsible. Right. I don't think that people want to break laws and hurt people by and large. Right. But when we make it inconvenient versus the things that they need to do, everybody's got a ton already. we make it inconvenient to govern, be responsible, comply, then they're going to take the path of least resistance. And so I just wanted to lower some of those barriers. Do you see an environment where people are still resisting governance and they feel that, you know, it is difficult for them to, for it to be irresistible? Of course. Yeah. Because, look, our product is not the only product that exists in this space. And I'm sure that others who pursue these spaces are attempting to do something that is as intuitive and beautiful and everything like I just described. But at the end of the day, this is not just a technology matter. This is a process matter. This is a people matter. So this goes back to those three questions. For what purpose? To what end? For whom might it fail? Those have to undergird what I refer to those three questions as ethical inquiry. They have to undergird, really, all of the technology and process and so on. And so there's got to be a will is really what I'm describing. There's got to be a will to pursue governance in a way that is, of course, necessary. But we've got to be able to distribute that will beyond, say, the compliance team. So you just said it. Technology at the end of the day is an enabler. And when we look at the transformation we are going through right now, it's about people and culture. So during your keynote as well, you argued that AI governance is about preserving human culture within organizations, not taking the regulation or compliance box. Right. And we see that happen often in large enterprises because there's so much to get done. That governance is actually a system for scaling judgment. And I think judgment is very important so that we can preserve that culture. So why does that framing change how an organization should approach it? So if we want people to take things like ethics and responsibility seriously, we have to be able to speak to them in terms that matter to them. Right. We can't browbeat them because that has extremely limited impact. Right. We can't just appeal to their better angels. And quite frankly, we can't just appeal to the pursuit of profit either. And I think that's a bit of a misnomer. But I think global appeal becomes really important. And one of the things I mentioned during that keynote was this notion of culture. And that's because it's something that all of us really can identify with. Like we're all a part of culture. It's what gives us a sense of identity, gives us a sense of belonging. And that's also true, not just at the macro kind of country level, it's true inside of organizations as well, because organizations thrive. The best organizations, anyway, thrive on connectedness and cohesion. Now, ask yourself what happens when employees rarely come to the office, as an example, or if training is recorded and on demand. And at the same time, we're outsourcing chunks of our cognition. Maintaining that connectedness becomes extremely difficult in that setting. It gets really, really hard. And that's why I say governance really is about scaling our judgment. It's very difficult to do that when there's disconnection, right? When there's a lack of cohesion. What I also said in that same speech is if we want to preserve our cultures, then we've got to be able to preserve our judgment and vice versa. Right. This is why governance becomes really important because governance is a way of scaling that judgment. Right. And so, and we can explore, I know it gets a little philosophical and esoteric, but the general idea here is culture is a reflection of us. We create systems to maintain cultural norms. But if we aren't careful, the systems that we create into reshaping the very culture because of the needs to comply with the terms of the system. And so it's extremely necessary at moments like this when we talk about the foundation shifting. As we are shifting the foundation, albeit in some cases intentionally, in some cases unintentionally, we've got to be extremely cognizant around this idea of embedding necessary human judgment so that we preserve necessary human culture. Yeah. You know, as I listen, I also I'm thinking about some of the conversation I've had with executives in recent weeks. So one of the conversation, you know, we talk about governance and we also talk about ethics. And one of the questions I realize is, you know, as I, you know, like many of us, you know, I had my jobs and I work with great companies. ethics in some ways are imposed on us by corporation. And I'm going to try to explain where I'm going. So, you know, you have rules and way of behaving, things you can do, things you cannot do, right? Which are considered as ethical within organization. But as we are moving to this new world, I realized that ethics needs to be part of who we are and how we decide to use, For example, AI systems. You know, there's a lot of shadow AI happening in organizations just to help us do a job better and potentially to still keep competitive advantage, right? When there is so much fear around what AI could do for us or maybe take away from us. And this is because of the framing. But I realize ethics really sit into how we've been educated, into culture and what we believe right or wrong is. The reason I'm saying this is just two days ago, one of the big insurance companies put an article out around AI fraud. And now people would go onto the internet on Facebook and actually see whether you have put a picture of your car with the registration number. They are going to take that picture and then actually put fraudulent claims. And so you think, OK, is, you know, is that ethical? You know, I share my great car because, you know, some people love sharing their great belongings. But actually, I've created a problem for myself. Somebody is going to take that picture and somebody is going to create an accident and damage to that vehicle and then put the claims forward. So what I'm trying to say here is when you think about governance ethics, you know, when you think about the way things are being built around us and then culture, what does an AI system stop reflecting when you think about our value and start producing if they are producing new ones? This is a great question and even better dilemma that you just posed, because what you just touched on is kind of the layers of ethics, right? So we started with this idea of corporations having policies effectively and how the values of the corporation are embedded in the work ethic of that organization. No two organizations are the same. And so those ethics may change. But those ethics then are subject to a larger ethic that is more of a, say, a country level. Yeah. Right. And those ethics get embedded in country level policy that we call law. Right. One supersedes the other in most nations. Right. And so what you just described in the dilemma is what used to be part of. And let me do one slight backtrack. And I think ethics, I'm thinking about a key term, social consensus. Right. And so what do most of us believe is the best appropriate standard, right? Those sorts of terms come to mind. So if social consensus at one point in time was show your belongings, this is great. No big deal. Everybody's having fun. This is my cat. This is your cat. But now all of a sudden new technology comes along that makes prior social consensus a vulnerability, which is what you just described. So now all of a sudden we've got a fraud scenario in this present day. Now, guess what? We can't go back and reclaim all of the car picks that we put up, right? They're there. And now people are taking full advantage. Now, in the case of the layered ethical approach, they are still breaking the law. That's why we call it fraud, right? Now, the next step is enforcement. How do we deal with that, right? Another step is things like figuring out ways to remediate before it occurs, if that makes sense, or to mitigate is probably the better way to describe that. So to keep that from occurring. So now that puts more onus on the technology providers to say, wait a second, I see you taking a picture of a registration. We don't allow registration pics, right? We're dealing with some of that with some of the latest model releases, you know, Fable coming out this week. You know, if you start asking your biology questions, they're going to flag it, right? Because now that guardrail is up because of the present moment that we're living in. And so when I go all the way back to where we start this conversation about the dynamism and the fluidity, this is what I'm describing, right? There's a lot of foundational pieces that are changing right now, right underneath our feet. And we have to accept this space. I think we also have to, because I can hear someone thinking, well, just regulate it, right? It's easy to say it's a lot tougher to do at this pace of change. And so we just have to accept that there are some things that we're not going to be able to just stop. Doesn't mean that we shouldn't try to mitigate the harms associated with them. But I do think a prudent approach is to allow us to reach some point closer to homeostasis than we are right now. So I'm doing some work around AI trust deficit right now. And one stat, or actually a couple of stats I just saw last night, which I would like to flag to you is Gen Zs, 55% of Gen Zs, think that altering a picture is OK. Millennial, I think, is around 29%. So it's about value, right? Again, we are going to believe, right? So in the industry I serve, insurance, right? You pay your premium and you're expecting as you paid your premium. So that is psyche, that if something happened, you will be reimbursed for what happened. And I think when I think about Gen Zs and millennials, it's 55% and 29% said, okay, well, I lost, I mean, I damaged my phone and it's a small crack. Well, let's just make the crack a little bit bigger, right? Altering the picture. so that I can get a new phone, right? This is potential sake, but I paid my premium. So therefore, this should be okay. And so what you find is that misalignment between what is rules in society and how, I would say, an industry works and what is done. So a number which came about is 46% of companies have an AI trust deficit right now, leaving half of their AI potential on the table. So what does that trust deficit actually look like inside a company, Reggie? So in anticipation of that question, I wrote something down here, and I'm just I'm literally going to to read it to you, if that's OK, as this came out of our data and AI impact report. So nearly, as you said, nearly 46 percent of organizations sit in this so-called trust dilemma. And I just want to describe it for folks who may not quite understand what we mean by that. So they either underusing reliable systems because the confidence is too low or they over relying on unproven systems because confidence is too high And so what the study really got into was this notion of effectively a chatbot experience and how that is that feels much more trusting than it should be right? The technology that undergirds it is highly probable, say for a couple of guardrails and that sort of thing that you put up, but the heart of it in neural network is a probabilistic system. And so one should question whether or not we should put, you know, how much trust we should put in that. I think this is trust on range as opposed to trust as a binary. And so that makes that sort of technology, that sort of system acceptable in certain scenarios, less acceptable in other scenarios. So for instance, if you need to do something that is say health related, financial services related, which is more of your space, then you want to generally have something that's more deterministic. Like when you open up your bank account app, you don't want, you know, yeah, you probably have a thousand euros, like you need to know for sure. And that's probably a poor example, but you get my point, right? Some things you need to be deterministic about. So day-to-day in organizations, it looks like two scenes. So scene one is fraud. A fire fraud team has a validated, well-governed model, and analysts quietly route around it because nobody helped them to trust it. Scene two would be someone paste generative AI output straight into a board deck because the tool sounded confident. So this is kind of the overconfidence ratio inside of these generative systems. And then the stat that I found for you, because I know you like your numbers, 78% of organizations claim to fully trust AI, 78, but only 40% have invested to make their systems demonstrably trustworthy through things like governance and explainability and safeguards and all this sort of thing. So there's a gap there between our perception and the actual practice. And that's where that roughly 50% of the potential stars will leak out. So we got a new edition of that particular study coming out later this year. So I'm really curious to see where it goes now and how those numbers shift. But I think it's pretty interesting that, you know, so many of us say we want trusted systems. far fewer of us are spending the time and effort and process and tech and whatnot to put those kinds of systems in place, while at the same time investing probably an abundance of trust, an inappropriate abundance of trust in systems that we truly shouldn't trust for certain scenarios. Again, trust on a range, not as a binary. So deterministic versus probabilistic, I think this is very important for everybody to understand. So you would expect deterministic AI system to play, I guess, for your fraud, your bank account, credit checks. Probabilistic would be our generative AI, chart GPT. And even I, you know, when I do my research and I use a lot of deep research with Gemini, for example, I'm looking for my links and I go and check every single link, every statement. So there is a way to control that probabilistic system by having the ability to go and check out citation, I guess. This would be a good example, wouldn't it? Yeah, but that use case for research is different than the use case for determining whether or not we should release a drug to the public. Absolutely. Right. And so in some cases, one plus one always has to equal two. So in research, if you don't get every single primary source, that's OK, right? Maybe you get a preponderance of evidence and you go with it. But in cases that are generally high risk, you know, we think about the EUA Act for an example, for those high risk sorts of scenarios, you usually want to be precise. And that's the difference between wanting to use a deterministic system versus a probabilistic one. Yeah. And thank you for that clarity, because I think often when I speak with people in general, I think we go back to that deterministic versus probabilities right now. It is one I've noticed a lot. your sharpest architectural point was that company govern at the wrong unit they track and deploy red teams models when they should be governing use cases and i remember we talked about use cases versus models versus frameworks when we met at sas innovate so why does that order of cooperation change everything about how governance actually works. Yeah, because use cases are where value and risk actually live inside the enterprise, right? Not models and agents and so on, just on their own. So generally speaking, models and maybe to a lesser degree agents, they have a broad set of abilities, but how people use those models and agents are usually fairly narrow. Go back to your space. You know, you think about loan adjudication or claims triage. If you're in HR, resume screening, those sorts of things, you can use the same model to address each of those particular use cases. But each of those use cases have different levels of impact, different levels of risk, different levels of value for the organization. And so the risk in this particular case and the value belong to the use case and not the engine. I think you brought up electricity earlier as a parallel. It can power everything. But what we really care about in this studio right now is whether or not the lights are on. If we were engineers for the local power company, we would probably care about the big transformers and that sort of thing. And so the electricity itself has different impacts, different risk, depending on where you are in that value chain, if you will. And so that's all we're trying to get at with this notion architecturally, which is if we focus on use case, we put people in a better position to deal with the things that they care about. If we go back to our making responsible, irresistible, people care less about, you know, you look at all the benchmarks associated with the models and whatnot. People care less about that at the point of impact, right? It's cool for people who are in the model building business, but it's got to be translated down to the use case for you and I. And so that's really what we're attempting to architect around. So now when you think about the CEOs of companies, right, when a CEO tells you, you have got governance handled, we audit all our models, Reggie. What do you say back to that? I say, show me your work. I would say, first, auditing models is necessary, but it's insufficient for all the reasons that I just described to you before. And I would also say, show me where you believe harms are showing up as well as value is showing up inside of the enterprise. But at the end of the day, it starts with things like just the inventory, just start with the inventory. And then from there, we can ripple out into some of those other conversations. Well, I guess that takes us into AI Navigator. And I had the great opportunity to view, test the platform with your team. And now we just- Honestly, what'd you think? I loved it. I mean, it's beautiful. Just tell me that. I mean, you know, as somebody who come from a process background, right, I did a lot of process transformation. I could actually very much visualize, you know, what you do. I mean, now you have Fable Cloud that you can upload into it. And actually, I can see organization being able to test what is coming up out of the work they are doing, being able to flag things which might not be, you know, within the governance and ethics of an organization. And so, you know, when you start seeing, you know, how those AI use cases, then we can look at the framework, we can look at the models, the agency, maybe the policies, you know. So what are you learning? I mean, how are you learning to build it, you know, where it is now? Because I know the team is working as well with the customer to make sure it's meeting the customer experience requirements and how they can become better at making sure that the organization is safe. Yeah. It's an ongoing journey. You know, I think, go back to my earlier comment about certainty. Certainty creates a measure of stability, but it also, you know, creates stagnation, right? You can't separate the two. So there's some things that we have to be certain about in the product. What makes the product minimally viable has to be certain for us. And so we're marching toward that. We can't be certain that we have captured everything and therefore we have to have all the conversations that you just described. We also have to ship, right? And we have to honor the commitments that we've made at the point of shipment while at the same time continuing to innovate. Now, that's not new to us and our product development cycle, but that is the current journey. And why that becomes important is that we want to make sure that it continues to reflect present and future needs of our customers as best we can. So I hope we get that right. Yeah. Yeah. And you also said that the biggest risk is under regulation as part of that frame. A tool so complex, no one use it, right? We want people to be able to love it. It has to be resistible. How are you designing Navigator so it is, you know, really embedded in everything? people do every day. It just becomes maybe like we open the newspaper, right, every morning, the things we actually check in, first things when we walk into the office. Yeah, well, so I'm clear, engagement is not the goal here with this product. You know, exercising good judgment is. And so the way I like to use it is I want to go in. And so one of the things that I told the team when we were first starting was I need to know how, quote unquote, how we're doing. Right. Boss always walks in the room and says, OK, guys, how we're doing. Right. I want to answer that question first thing. Right. And so if you look at the upper left corner of the homepage for Navigator, it says, how are we doing? Right. And then it gives him a frown face, smiley face, that sort of thing. So they keep they keep it cute. But if it's a smiling face and I've got nothing to do, shut it down and go and do something else. You know, it's not there for me. Well, now I have a certain persona, kind of an executive persona. But if you have an operator persona, you may need to go and dig, right? OK, I see smiley face, but I also see we have a couple of tasks. And so let me dig into the workflow. Let me see, you know, if this model is still, you know, within range and all those sorts of things. but I want to spend five minutes or less. And that I think should give hopefully some measure of comfort to people who are in roles similar to mine, who are not looking for yet another window to have to go and explore and spend time with. You know, we have to recognize, and this goes back to my irresistible comment, we have to recognize that there's already enough on plates today. We don't want to add to that. In fact, we want to find ways to take from that so people free up time. And our hope is that by us doing some of this background work of not just inventorying the use case, but showing if that use case is tied to certain models or certain agents, making sure that that use case is compliant with a certain set of policies within the organization, making sure that it's compliant with EU AI Act or what's going on in New York or Colorado. or what have you, and have an ability to connect to your financial systems to tell you how many tokens you use. Like, we should be able to just know that at a glance and then make a decision about it. Yeah, absolutely. Right? As opposed to spending, you know, the next four hours playing around with something that SaaS builds. So that engagement is not the goal, right? And extracting good value, exercising good judgment, those are the goals. And so, interesting enough, right, I'm thinking about our compliance or governance officers. I mean, do you see the world changing from compliance officer to governance officers, actually, Reggie? Because now I see a lot of chief AI officers. You know, when I look at every organization I work with, there are chief AI officers everywhere. Yeah, everybody just got a promotion. Just extended their careers. I don't know. Maybe. Maybe. You know, I think it comes back to the type of enterprise. We have, say, compliance reports up through me. Will they continue to be compliance? No. Personally, I didn't favor the term. So let's recompose this. But in that recomposition now, we expanded breadth for the team to think about the standards and regulations that deal with compliance, but also to think about it from a risk intelligence perspective, which is a much more, if you will, maybe this is more of an American term, but putting us on our front foot as opposed to, you know, reacting on our back foot all the time. I think nobody wants to see the compliance person walk through the door, right? But maybe people are much more interested in the person who can give them some intelligence around the vulnerabilities that they have vis-a-vis risk. But that's probably more of a philosophical organizational question. I do think that in some organizations that has to persist, right? because it's part of the normative behavior of those organizations. But that's okay. I don't think that has to, you know, that naming convention doesn't have to determine a ton there. I think what's most important, though, is burden and where the burden lives inside of the organization for compliance to standards, regs, not just externally generated standards and regs, but internally generated. There are more internal policies in most places than there are external regulations that they have to be concerned about. And so that burden, in my view, organizationally shouldn't live just on one person's shoulders, one team's shoulders. That should be a distributed burden. But that necessarily means that as we distribute some of that burden, that we don't create additional hurdles, right? And that's the more difficult thing to do. It requires us to think about standards and regs at the point of design when we're thinking about process, when we're thinking about technology. And unfortunately, we don't all have that luxury because we've got to go retrofit now. We're talking about organizations that have survived for decades. And now here I come with this great idea about distributed burden. So we've got an opportunity now with all this foundation shifting to go and re-explore some of those areas where we might be able to distribute some of that, quote unquote, compliance burden. But I think that window is going too close. And so now is the opportune time to take advantage. It's interesting. I think I want to unpack this because five years ago, I was doing a lot of work around sustainability. And if you remember, sustainability was pretty much a checkbox. exercise. You had somebody doing sustainability and they had to report back. And gradually, what I saw is that sustainability role became a board level role and often actually reporting into the chief risk officer, chief finance officer. And that is what has happened over the past five years. And I love the way you are describing compliance. You know, it may be from pain to pleasure, from compliance, we are going to move to risk intelligence officers. And so from burden to pleasure, I mean, what could that be? I don't know how it looks like it's pleasurable, but... You know, think about that role again, moving back to, you know, becoming a board level, not a tick box, exercise, right? A real board role, valued accountability, responsibility. Because what I see with artificial intelligence is actually the role is moving to the chief finance officer from a risk and proof point, you know, investment viewpoint. The CFO is the one, however, to do a lot of work, work he probably or she didn't have to do in the past. So I'm thinking about Monday morning, right? Making responsible, irresistible for our chief risk intelligence officer arriving in the office on Monday morning. What should that look like? Oh, well, I'd like to say they should just open up Navigator and it does the work for them. You know, I don't know that the measures and the metrics and that sort of thing necessarily adjust for these individuals. But I do want to go back to this notion of the distributed burden. I think what Monday morning should look like is less pressure to provoke people to do what is in their best interest. And I think a lot of when I have conversations with folks like that, they have a tough time understanding because they see their role as one of helping to protect the organization. They have a tough time understanding why others don't share that same commitment of protection. And I don't think that it's that people don't want to protect the organization any more than the compliance person does. But I go back to they've got 17,000 things to do already. You're adding one more and it's not quite clear what value it adds vis-a-vis the goals and objectives that they've already been provided. I don't know that a compliance officer is ever going to win that fight. And therefore, if you embed the needs of compliance just into a natural workflow, it now becomes part of achieving the objectives that that person has been delegated to perform. And so hopefully the idea of distribution can really resonate here. And I get it. Not every organization is structured, has the leaders with the political and moral authority to pull it off. I get that. But I think that has to be the aspiration, because when you centralize this notion of compliance and governance and safety and so on, it becomes the job of an individual eventually. Right. But it has to be a collective action. Yes. And I think that's where I would want to leave that answer is do all we can in pursuit of collective action. So we have a shared responsibility because we're all going to have a shared reward and a shared risk. Right. So why not share the responsibility as well? so i would like to move into a term a transformation which is happening to us right now which is this agentic frontier transformation so let's put some definition when we think about the frontier transformation is actually crossing the frontiers like going from one country to another, right, is that world where we have access to intelligence on tap. Today, we know that intelligence has become pretty much utility. One million token is less than one dollar. I mean, you can cut it in different ways in the fact that also companies realize it's very costly to use new intelligence. But this is more than that. It's about human and agents working together collaboratively, like on the AI navigator, right? I have my smiley face. I have my agent working in the background. I'm getting insight as to whether they are doing a good job or whether I need to actually be, you know, providing some warnings because there's a bit of allotination happening here and there. But it is also about each one of us taking accountability, responsibility for the agents we are working with. So becoming managers. And I often say to interns now, which are coming and knocking on my door. You know, you as an intern, you want to probably, what we used to do in the past, do some research and go and find the reports. No, companies do not need you to find reports. They can find the reports themselves using agents. They need you to be manager. They need you to be creative. They need you to apply judgment. They need you to be a system thinker. So I would like to go into the agentic frontier discussion because today in many enterprise machine identities, and I think you will need to explain what this is, outnumber human employees and most companies admit their agents have already done something risky. And that is a thing everywhere. We know there is something which has happened. I've seen some consultants and I think when you look at the consulting reel and some of the reports we've seen where, you know, reports have been released where, you know, factually inaccurate things is because, you know, we work on projects together, many of us, and we think somebody else has done the checks. And I think that is what happens often in big organizations. Somebody else has done the job. So when an Asian acts and no human told it to, it's gone and do things we didn't ask them to do. So who is accountable with that? Human. Right? Yeah, it's not even a question. Look, humans are accountable for what humans create. Artificial intelligence, whether it's, you know, the frontier stuff and models, whether it's agentic, whether it's deterministic, whether it's quantum. whether it's digital twin, you name it, we create it, we're on the hook. Full stop. I think to suggest anything else would be to obfuscate our duty. This is our thing, right? No different in cars, no different in lawnmowers, right? If we create it, we are accountable for that creation. And what we should not do is to scapegoat the technology. What we should not do is attempt to dodge the repercussions of our own choices. Right. And I think why I'm so adamant about it is, you know, this is a legit conversation, by the way, what you're bringing up. But where it goes is, you know, the legal identity for AI and whether it should have personhood and all that sort of stuff. No, stop. They're not human. Period. We created it. We are on the hook for it. And we should again not obfuscate our duty Yeah Right Importantly if we go down that path we effectively strip ourselves of our own agency And that's a path toward handing the designers of current AI far more power than they already have and certainly more power than they deserve. Yeah. I don't want that. I don't know about you, but I'm not here for that. So I think we have to kind of seize the moment and stop people in their tracks when they start down that path. All right. An agent is acting with human instructions. It is. We have created this probabilistic technology. It's doing what it does. Yeah, we can go into whether or not we know we're exercising judgment alone. I get all of that. But inside of that system is some measure of human judgment. and we should not entertain the ideas that it acts independent of our choices. Whether we are designing it, whether we are deploying it, and whether we're using it and everything, right, along that spectrum, these are our creations and accountability has to follow our choice. So what does human in the loop, on the loop, out of the loop, means when you have probably 10,000 AI loops running in the background? I'm a woman and human. You know, how do you manage 10,000 bloody loop in the background? You figure out a way to govern 10,000. Or if you can't do that, you stop until you can. Okay. All right. Because we have to maintain some measure of control and accountability. You go from 10,000 to 100,000 to a million. And now you just have, you know, agents running around making, what, decisions on our behalf? Yeah. Like, is that the world that we want to inherit? Again, I go back to for what purpose, to what end, and for whom might it fail? We have to be able to answer those questions. And if we can't answer those questions, we should probably slow down or stop, right, until we can. But this notion that it's all inevitable, it has to keep going because it keeps going, it feeds on itself and there's nothing that we can do about it, is a false notion. It just is. So we should take control, but it requires competent leaders, courageous leaders. And right now we have a shortage of that. I'll be just frank. So we've got to make sure that in the absence of that kind of leadership, that the rest of us say, wait, hold a second. Let's vote the right people in that will be competent leaders, courageous leaders, conscientious leaders. Let's not help continue funding the operations of people who don't reflect those kinds of values. Right. Like this is a collective action we talked about. This is a collective action matter. And none of us have the luxury of sitting on a sideline right now. Right. So one of the reasons that I go out and I speak as I do, travel the world as I do, talking to people as I do, like, you know, hopefully this gets platformed, is because AI literacy is probably one of the greatest necessities around the globe right now. I go back to your electricity example. I often draw the parallel there. Most of us aren't engineers. Most of us don't know the nuances of power generation. But I'd be pretty comfortable in saying the overwhelming majority of us know how to plug something into a wall and get what we need out of it. Yeah. Right. That's that requires a foundational level of knowledge. Right. If you will, a common sense. Right. Yeah. And so we need to get to a common sense when it comes to the topic of AI. Not everyone will become a data scientist. Not everyone will become an AI researcher. But we all need a common sense around some of these things. Why? Because the common sense helps to inform some of the choices that we want to have in our lives. Common sense around electricity helps you understand the value of refrigeration, helps you understand the value of lighting. We need more of us understanding some of the common problems, the foundational problems that lead to conversations around data center buildup. that lead to conversations around mental health consequences from, you know, extensive chatbot interaction that lead to conversation around governance and fraud. Like, oh, you know, we're having a sophisticated conversation, but I'm talking about bringing it way down. This is a data literacy conversation as much as it is an AI literacy conversation, but we need that common sense. And this is a global phenomena. And I suggest it not because of the opinion that people need to use AI. Quite the opposite. I'm approaching it from the standpoint of enabling people to make decisions about the choices that they independently would like to have occur, right? And you can't do that when you're not sufficiently knowledgeable about a topic. And so I just want to raise the level of competence so that we all have a measure of fluency. Maybe not everyone will become completely fluent, but I think you see where I'm going with it. Like, we all need to have some conversational knowledge about this topic that's not rooted in, it's going to kill us and, you know, all of this, or it's going to create some utopia that, you know, is on the other end of that spectrum. And so, yeah, AI literacy, I'm all in for it. And it's a personal passion project. Yeah, it takes me a little bit to the trust by design conversation I've had also with some of your colleagues, because one, I remember one of the statements during that conversation was about the fact that what you are saying, right, businesses are building a lot of agents. And then there's this expectation that the human is going to monitor and actually manage those agents. But no human is able to manage 10,000 agents, right? And it's where, I guess, having some level of ability to govern, evaluate, and actually understand what is that human-agent ratio is going to be more and more important. In some roles, it might be, I don't know, maybe 50 agents per human. But for some roles, it might be three agents or maybe one agent. And partly when you start looking at highly complex commercial line claims where you're human, the empathy, the judgment is so important. Which takes me to my next question, because you told me governance is a system for scaling our judgment. But agents scale action, not judgment. So are we scaling our judgment right now or quietly outsourcing it to technology? Yeah. Yeah. So let me say, based on the way you frame that, I'm going to challenge the premise that agents are scaling action rather than judgment. OK. I do believe, though, we are outsourcing some judgment, just so I'm clear. So I'll meet you halfway on that. The action that agents are taking are really a function of judgment. Sort of the conversation we were just having a moment ago. For instance, these reasoning models, right? They are attempts to approximate human judgment, if you really think about it. But human judgment itself is also an approximation. It's an approximation toward some truth, as we would describe it in language. But human judgment is fallible, important to know. And human judgment requires inquiry, discoverability, some measure of assessment against what we might call a standard. Now, all of these are imperfect creations, though. So I'm going to take that notion and put it to the side. Now I want to compare that to, I'm going to get really philosophical with you, compared to like divine judgment. Right. The term that we see show up in religious scripture, Bible, Torah, Koran, you name it. This notion of divine judgment presumes a complete truth is already known. Right. So I apologize in advance. I dig in the historical context and etymology because I think words matter. Yes. And the essence of language matters. Right. And so that's that's kind of why I go here. So go back to this divine notion of judgment presumes a truth is known. And then judgment in that context really is about determining a fate. If you are judged by, you know, from on high, right, that sort of thing. Right. My concern is that if we over-index on agentic systems, I think we run the risk of pushing our culture toward more of that divine version of judgment because it's an external judgment. It knows all, right? There's a certain level of omnipotence that's associated with that divine type of judgment. As I go back again to the human judgment, which we accept as has some failings. Perfect example, just to try to come out of the clouds for folks. If you go to a judge, let's say you've go back to your fraud example. You've committed fraud. You've been found out. You're one of the people who took the registration and did something wrong with it per law. Right. Judge says, Sabine, you've been bad. We're going to fine you 10,000 euro. Have a nice day. I come, Reggie, you've been bad. We're going to jail you. Same offense. Same offense, but different outcomes. But different outcomes. Yeah. That suggests there's some fallibility associated with the human judgment that we exercise through law. In our notion of this divine judgment, there is no fallibility. It's completely infallible. There's a certainty associated with it. And so when we start to think about that external certainty, what are the genetic systems? They're external to us and they're just creating outcomes. Increasingly, we're starting to give them the ability to make decisions absent our participation. How far are you willing to let that go? For what kind of case are you willing to use that? And as you continue to trust that more and more and more, you inevitably create what's known as an automation bias. Automation bias is great if it makes sense for you. I trusted Google Maps to get me from where I was to come over here. I have used Google Maps in other scenarios and it's like, are you sure you want me to walk through this dark alley right now? All right. So now I have to exercise some human judgment and say, I probably should go to the lighted path. And so my concern is we think about culture and we've got young people coming up. And if they are going to be inheriting these capabilities, we need to also embed in them the ability to think critically about these kinds of topics, which goes back to my literacy. because if we don't, they're going to lack the ability to know that they shouldn't walk through that dark alley. And as the foundation is shifting and people are, you know, doing new things because of capabilities like your fraud example, someone's in that dark alley waiting on them. Right. And so, you know, it gets a lot deeper than just some of the some of the words that we use. I think, again, the essence of the language matters here. And what we're truly trying to accomplish matters here. We've got to preserve human judgment if we want to preserve human culture. I completely understand that. And going back into judgment and the example you just used around younger people, you know, being comfortable, for example, to acquire knowledge through, you know, a lot of social media platform, which, you know, a lot of misinformation, disinformation, which we've seen even in finance, in insurance, having impact on mental health, of environmental risk because of misinformation actually preventing people to actually apply personal judgment on things. And so it's important to go back to basic, not being lazy, you know, to actually look at different sources of information, making sure that we know that something has been validated by authoritative, for example, sources, rather than just taking off information from, you know, social media platform as an example. That's what we would, I'm going to date myself, back in the day, we would just call that research. Yeah. It was basically, you have your primary sources, you have your secondary sources and so on, and you want to make sure that you're, you know, comparing them against one another. But increasingly, that skill set is being lost because of the success of creating accurate systems that we believe are reliable. And again, that makes generally makes sense if you are doing a research paper for middle school. That makes a lot less sense if you are doing research for a prescription drug. Right. Just different scenarios. So, again, I don't want to poo-poo the use of AI, but I do think it's important that we understand the proper use cases for AI and that we treat them with the level of rigor that they deserve. Yeah. I mean, you know, even I think in recent, you know, weeks and months and, you know, because I use social media platform as a communication educational environment and I trust that they can actually provide education as long as the ethical human put valid information out there. Some of the content is so realistic. And so, you know, you have platform now like X having those notes. And sometimes I look at the notes like, oh, my God, this is so realistic. And you need actually that validation and that double checking to actually make sure that what you see is accurate. Now, more and more, more so than we have ever needed in the past. Yeah. Yeah. People should recognize that this is a first for human history. Right. So this is a fascinating time to be alive. For the first time in human history, we have either the privilege or the curse, however you want to look at it, of not quite knowing if we should believe what we're hearing and seeing. And we should pause on that. Again, I go back to the dynamism. We should pause on that point and question whether that's the world that we want to live in going forward, what that means to us as a species. What other species on the planet can't believe its senses? Right? What does that do to instinct? Are we comfortable with that? Right. Again, I'm not making a right, wrong, good, bad statement, but I think it's fair to ask the question so we can determine, OK, now that we've got this, what do we want to do with it, y'all? We should probably put some guardrails up in certain cases, right? So you look at what the EU is doing with the AI Act, the transparency rules, Article 50 stuff. They're basically saying, you know, look, if you're going to create synthetic data, part of the transparency requirement is let people know. If you're going to have the synthetic data interfacing with the public, let people know that it's synthetically generated. To me, that seems like a good idea when you especially think about it in the context of, I don't know, things like wars and whether or not missiles are falling out of the sky for real or if it's just some animation. You know, I do work in NENA right now. And, you know, what happened in the region, for sure, uncomfortable, but there was a lot of misinformation. And, you know, the Emirates, you know, Dubai government actually highlighted that they would arrest and they would actually pursue people putting misinformation out there. But, you know, because of AI, it's easier to actually create fear. Let's go into the 2nd of August, right? We have a few weeks when we look at Europe with the EU AI becoming fully enforceable, the first binding comprehensive AI regime that anywhere would have seen. And so we have a hard deadline here in Europe. And so what does that mean, you know, when we look at our goal post and, you know, the things we need to consider and do to be prepared as enterprises? Yeah. So let's update the audience. So the August date is still a real date in terms of beginning enforcement, but they push back the high risk obligation to December of 2027. And that's just for the high risk stuff. But like this Article 50 stuff that we were just talking about around synthetic, they move that to December of this year, if I'm not mistaken. I have to go back and verify that for your audience. But nonetheless, that's not to suggest that that is going away. I think that's more of a recognition. It would take time. That it takes time to do some of these things that are being asked. And so I would suggest that folks see that little reprieve as an opportunity to continue what you're already doing. And if you haven't started, you better get started as long as you are either, you know, a business inside of Europe, the EU, let me be really specific, or you're a business who wants to do business in the EU. All right. This isn't going away. And so we should do all that we can to to ready ourselves for that. And so that's certainly what we've been doing at SAS. You know, I've had the pleasure of working with a lot of folks on that topic. And, you know, knock on wood, we'll be ready. And, you know, we've been participating in the EAI pact. We're one of the first signers onto that. And so we've had the privilege of kind of being tapped into the information flow. So I'm extremely thankful for that. But by and large, I understand the detractors' points of view, but I also understand the advocates' point of view. And if you force me to place a bet, I would say I'm going to lean in the direction of the advocates on this particular topic, simply because it is an attempt to go back to those first set of questions, for what purpose, to what end, for whom might it fail? And if we can foresee the failings, I think as competently as we have a responsibility to do something about it, there will be things that will occur that we cannot anticipate. But look, welcome to life, right? But for the things that we can anticipate, for the things that we are already seeing, like your Dubai example, I think if we want to preserve what we understand as human culture and society in general, and I get society doesn't work for everyone right now, but let's use the capabilities to expand what society should be so that there's more participation. If anything has come out of all these capabilities is a democratization to intelligence, which I'm all for. And so let's figure out a way to maximize that effect while at the same time minimizing some of the foreseeable harms that come as a consequence of some people who are attempting to profiteer and take advantage. Let me stop it there. I mean, you talked about the fact that we talk about the EU AI Act, which is about European Union, and the UK often follow on, right? They actually tend to actually agree that those laws need to apply as well in the UK. But when we think about other countries, right, what about international markets, companies, maybe in the United States, maybe in Asia? You know, what does that EU Act mean for them? Well, again, if you want to do business in the EU, I mean, it's a pretty significant economy. And so if you want to do business in the EU, then you're going to have to comply with the laws of the EU. Just like you want to do business in the United States, you have to comply with the laws of the United States. You know, they say trust travels through the supply chain. And so if you sell to buy from or have the need to process data in Europe, then you're going to have to comply with these rules. I mean, I don't know what else to tell you. There's a telling data point, though, that I think is pretty interesting on this particular topic. I mean, you look at the global surveys right now. They show that the EU is trusted more than the U.S. and China. you know, kind of the AI powerhouses, if you will, when it comes to regulating AI effectively. And that's the important word, effectively. Effectively, yeah. And that means that I think the EU's framework then becomes a reference point for customers and partners and so on who are looking at effective AI governance. You know, whether you like regulation or not, if you are a fan of effective governance, right now the survey suggests that the EU is doing it best out of the other economic blocks. Yeah, it's good to hear. So now let's go back to those tensions, because when we met, you actually described three different tensions that every enterprise need to, well, are going to face, but actually need to pay attention to. Moral alignment, operational efficiency, and financial value capture. Which one, you know, let's explain what they are and which one should organization focus on first? Should they focus on any one first, you think, Reggie? So let me define them for your audience. So moral is aligning AI to the organization's beliefs and values, just broadly speaking. The operational is using AI for efficiency without losing clarity and consistency. And then the financial was capturing value without taking on unmanageable risk. So I'm sure your insurance audience will identify with that. So I tend to work with sizable enterprises rather than, say, startups. So they don't have the luxury of seeing those tensions in sequence. Like they exist, right? They've been around for a long time. They're layered. They are oftentimes integrated, but I think they present more of an optimization problem than a sequential problem. And I think the best that we can do for now is think about it that way. So rather than me trying to situate it you know first second third for you I think trying to get the best out of each of those tension points and maximizing what an organization can do best for now is the way I would go with that response as opposed to attempting to suggest that one comes before the other Another thing too is industries are different Organizations are different. And based on where you are in the scope of the industry, where you are in the scope of kind of global geographies and all those sorts of things, you really have an effect on how much you need to lean into or against those tension points. Yeah, absolutely. Now, I want to get into the fact that you advise amazing people around the world, and you also advise the White House on AI policy, and you have done that for many years. You're also now building an enterprise governance product. So what are boards of director today still not doing that they really need to get done? I would say 24 months, but you may say, oh, Sabine, 24 months is really too far. 24 months is a long time in AI years, I tell you what. So just for clarity for your audience, I'm not in the position of advising the White House any longer. Okay. All right. Just get that point out. Jeez. I think from a board standpoint, I mean, let me put it this way. A lot of boards can be stagnant and that's not necessarily a bad thing. When you think about significant enterprises, what you can't do is overreact to every movement. What you shouldn't do is overinvest in, you know, certain areas. That's how organizations maintain stability. You have to understand where your advantage is. You have to pursue new advantage and all this sort of thing. But at the board level, the notion of when I use this stagnant, some might call it conservative, like that's a prevalent trait and understandably so. Let me just start there. I do, however, think I spoke to a few of them years ago when we were just kind of getting started with the practice at SAS. and they were stagnant to a fault. I had one director ask me, he said, you know, we've seen cyber, we've seen crypto, we've seen telco, we've seen, like, I mean, he was a longtime guy, right? What makes this any different? Yeah, yeah. I said, dude, this is, I think this is worth a reaction. I don't think it's worth, at the time, I don't think it's worth an overreaction, But I do think this requires a different level of investigation than what you're offering it. I think organizationally should probably start to dedicate a few resources. You should probably find out where it's already living inside of your company and maybe, you know, aggregate those resources. And so that was an example of one conversation, but that was replicated. I think if we fast forward to now, Sabine, you see the exact opposite, which is an overreaction in the form of investment, which is generated by the fear of missing out at this point. Some of the smarter organizations called on early and said, OK, let's figure out what works for us and pursue that. But those who are catching up now are just signing checks because, you know, somebody gave a proposal that said AI. That's very reminiscent of when you could make a billion dollars because you could spell Internet. Right. People were just. Oh, KYC? Yeah, people just throwing money, right? You know, there's a stat says three in four boards have approved major AI investments, but fewer than half of them have actually set governance expectations, right? And fewer than half have made AI risk a standing agenda item, right? Because boards function by agenda, right? You meet quarterly or whatever it is. But so the agenda guides a lot of the work. Only about 26% of boards discuss AI at every meeting. And among organizations reporting strong returns from AI, roughly 63% are putting it on the agenda versus 13% who are reporting weak returns. So let me say that cleaner for you. Those who are making it an agenda item and talking about it are seeing greater returns from their investment than those who aren't making it an agenda item. However, many of them are investing. So think about throwing money at a problem, not asking how it's going, right? And so you've got a pretty sizable chunk of boards that are doing that right now. And so that is problematic for them. And so I would hope that for those who are listening to your program that they are not in that gap, but I fear that many of them probably are if those percentages stand up. I mean, you know, the design of the organization has not changed, right? We are still org chart when we need to move to what we call work chart, right? We are working on projects and we are delivering those projects. We are moving to an execution world. And so, you know, the MIT reports around 95% of AI project fails. And we have seen the BCG that 80% of AI project fails. In insurance, we know that numbers will be between 20 to 30% if it's not more. And so what we actually see and what I hear when I talk to executives is, well, Sabine, we have invested 80%. You know, I said we invested in AI. But, you know, our budget, which was for 12 months, is already run down in three months because of the AI tokens and not having thought about the GPU architecture and all those things, which I find fascinating, actually. That the budget set for 12 months is already gone. Then when I talk to the young ventures, the startups, the scale-ups, I mean, a lot of startups which do not have AI native on the team actually are not doing very well because they're still doing what the things was fundamental in the solution they were provided to the market. Because everybody running for AI first, AI powered, AI native, when actually you think about the corporation, which is not an AI native corporation, right? It's probably going to become a frontier using human and Asian working together, but they are going to be AI enabled. And so there's so much confusion around all this language. So when you look at yourself, Reggie, and, you know, you meet a lot of those boards and you're advising a lot of those CEOs, what is the only things you actually give them? The slide, you are saying, listen to this, look at this, action this. What is the single thing to tell them when you meet with them? I'm going to sound like a broken record here. We like that because actually repeat means we are getting somewhere. Yeah, and for the younger folks in the audience, I'm going to sound like I'm on a repeat stream, right? But for what purpose, to what end, for whom might it fail? And in the case of just getting started, show me your use cases. Right. For me, that's the nucleus of all of this. You know, this notion of we want to be AI enabled, AI power. If you buy the argument that some have that AI is general purpose technology, it's just a normal technology like we've been using electricity as the parallel throughout this conversation. Computers, tablets. Can you imagine saying we're a tablet first company? Right. AI is a tool like a tablet is a tool. We're a computer enabled company. People say, shut up. That's stupid. Understand that it's a tool to be used in pursuit of filling your blank. Right. And then think about the use cases necessary to help you fill that blank. the order of operations in terms of business success hasn't changed. Yeah. Right. Just the tooling is changing. And so, yeah, I think people like to attach AI to the things that they're doing because they think they sound cool. And maybe they do. Right. Maybe they do attract a few more eyeballs these days. But to me, it sounds like the same. If everyone is using the same model, what differentiates your business? Absolutely. Absolutely. And that reminds me of a conversation I've had with some of the executives I work with in Australia. And we laugh a little bit about it, but hopefully, you know, beer with me is going back to Drucker, right? We were talking about AI strategy when actually we are still doing business strategy. And so I always wonder, you know, whether because we have those trends and, you know, we have those keywords and I do make the mistake every so often to actually get the point, you know, talking about AI strategy, which is actually about business strategy. who just happen to have an AI enabling component is why I quite like the word frontier, because it avoids talking about AI and really looking at operating model design, you know, how this intelligence is going to flow through that new operating model and how the business is going to be redesigned or maybe re-architected. You know, some of the processes may not look the same, but still serving a customer, right? A paying customer. Yeah. Look, the accounting equation, despite all the agentic and quantum and synthetic, the accounting equation has not changed. It's still going to be profit equals revenue minus expenses. Yeah. Right. And so, OK, what you do, the goal of the business is to make money to sustain the business. So we take it all the way back to the foundation, figure out how you work toward that equation. Well, let's talk to the CFO now, because when you think about governance, it's not compliance. It's about driving growth. It's a growth driver. So for the CFOs out there listening to this conversation, how does governing well actually unlock value for them, not just reducing risk or taking cost out of the operation? Yeah, that's a good question. You know, you mentioned the current dynamic of the impact of tokens and tokenization, the token economy, you name it. From a CFO perspective, what that feels like is variable cost, right? So in any business, you've got fixed costs, you've got variable costs. And any CFO will tell you they want to lean as heavily into fixed costs as they possibly can because they need to, as best they can, want to understand the cost structure of the business. So now you can build around that and understand what the revenue is required, right? What revenue is required. Well, when you have variable costs, you don't quite know where it's coming from. You don't know when it's going to hit and which quarter, what week. and you present yourself with a cash flow issue, and cash flow is king in a lot of businesses, particularly when you talked about your startups and scale-ups, like cash is king. I need the ability to make payroll. I need the ability to cover the needs of infrastructure and all those. That's the world of accounting and finance. And so the argument with respect to AI governance, I presume what you're asking is, We want to help lock down those variable costs as best we can. And we want to contribute to the fixed cost as best we can. And so if you really start to double click into that, I think there are kind of two terms I'll lean on. One is a financial term. The other is more qualitative. The first is cycle time, right? Because cycle time contributes to cash flow. So if I can go back to my use case, so if you can understand all the use cases that have a direct impact on your variable cost structure and your fixed cost structure, for that matter, you can prioritize the use cases using AI that are more contributory to either. You can conversely deprioritize those that might have larger variable costs. So let me try to make this more precise. If you've got engineering effort, you won't say you're developing a new product and you know that engineering effort is something that's going to require a lot of code generation. And you've got inefficient models. You've got young developers who are still coming up. Is that a project worth prioritizing given your current cast position? How do you answer that question today? Right. You don't effectively. So one of the things we try to talk about with Navigator is put yourself in a position to be able to answer that. So we can, you know, through cogeneration, you can simulate, you know, the number of tokens that will be required. Then we can put it in our stack of use cases and say, is that worth it today versus three months from now? Yeah. Right. So that's a cycle time conversation. The other is more the qualitative, which is trust when we talk about. So that was the risk reduction, as I just described it. The trust is more of the put yourself on the front foot conversation with the finance team, which is to say the same governance mechanism avoids the reputational harm associated with damages that are a consequence of AI being used in areas for which you had no visibility. And so if we can reduce the risk associated with that, the converse is we're increasing trust. And we can make a case about the positive impact of trust in the marketplace, right? Now, you may not be able to draw a direct line from trusted customer to, you know, dollar in Q3 or euro in Q4. But I think we can attribute value to a trust customer who trusts. Right. And so, you know, we can also impute value to those who are prospects who we want to convert into customers and that sort of thing. And so I think the right conversation with the financial part of the house is one around cycle time, cash flow, trust, and AI governance then becomes the enabler for those kinds of dialogues. I want to go back to this EU Act, EU AI Act. OK, deadline not August 2nd, but going to be December. But I want to actually echo to our listeners, which is absolutely critical, that they need to start early. I remember when DORA came about, you know, a year ago, you know, January 2025, you know, a lot of people like pull their hair and like, OK, I'm not DORA ready. The key things when I took to the regulator is now they are putting regulation for across the industry. The regulations start cross industry and then they feed into industries. Not all rules will apply to all industries, but what they are doing is now making it actually much more universal to apply low. So, OK, bold post, you know, has moved to December. But therefore, when we start looking at those organization enterprises out there, you know, what are the things they need to do today? right now to start getting ready for six months from now. Yeah. As you stated, everyone is affected slightly differently. So whether you are a provider, a deployer, or a user of AI, what the act does is speaks to each of those persona slightly differently. And so we all have different levels of responsibility based on our persona. Importantly, your persona may change depending on the nature of your work. So if you're receiving, say, AI from SaaS and you're changing, making changes as a part of your workflow, like we could be a provider. But if you take our AI and you do some things with it and sell to your customer, now you become a provider with a different set of obligations than you may have previously if you were just deploying. So the first order of business is just figure out where you live within that chain. The second order of business, of course, is to understand what the commitments are associated with that. Is it impact assessments, conformity assessments? Is it going to have a ripple effect on some of your measurement systems? Do you need to think about some things differently organizationally, workflow-wise, and those sorts of things. So I think most organizations hopefully are pursuing their interests in that regard already. So I'm probably not telling them anything new. I don't like to be alarmist about any of this, right? This is an externally imposed regulation, not unlike DORA or GDPR or any of the others, right? And obviously, Europe is very special, certainly special to SAS. I haven't done business here for 46 years, so we'll continue to do what we can to oblige the laws. Of course, we will attempt to guide where appropriate as well. But it's important to note that global organizations have this responsibility literally around the globe. So, you know, when I go to Japan next month, I'll be talking to the Japanese about their regulation and the things that they're thinking about. When I go to Australia, you know, yet the same conversation. So what's important from my view is to the degree that we can have it, some measure of consistency, which is why I've always been a fan of, if not a global standard, certainly a global understanding, which is when I was doing some of the White House work, part of the effort was to help build that network, that infrastructure so that we can eventually get to that kind of framing. you know, that has diminished a touch, but it hasn't gone away. So I'm hopeful that we'll, you know, come back around and, you know, that the United States will be a more fervent contributor to that conversation. But I think what would serve business as well. And, you know, certainly, you know, we've been around at SaaS for 50 years and we do business around the globe. Yeah, thank you. But not every company is that, right? There are some startups, scale-ups that are just getting started here in the UK or, you know, in Dubai or, you know, parts of Asia or whatever. And they need space in which they can operate and they need to have paths of least resistance to pursue, right? If I go all the way back to our early part of the conversation around governance, we have the ability as a global concern to be able to adjust. They may not have those same sorts of abilities, but the economy, the society has to work for all of us and everyone in between. And so, get off my soapbox, we've got to find ways as, again, competent leaders to construct the conditions necessary for more of us to thrive in our desired pursuits. small business, big business, everything in between, right? Individual citizen, billionaire, everybody in between, right? And so that is my ambition. Great ambition. I want to actually highlight a few things. As I was preparing to come and have this conversation with you here in the studio, Reggie, I was looking at who were our listeners. Our listeners from SAS, you know, Mexico, thank you. Brazil, thank you for listening. Japan, Thank you. And also, a lot of European, probably French government is linked to this. Okay. So, they are trying to work out, you know, how to use us better, I guess, but also Netherlands and Germany. So, now I have three quick questions. First instinct for you to finish our conversation. Okay. So, I would like you to finish this sentence. Okay. The companies that win the next two years will be the one that? Accept their duty and pursue it in ways that are beneficial to as many people as they might possibly benefit. I know that's long-winded and philosophical, but that's the first thing that came to mind. Love it. So one thing about AI governance, you have changed, which has changed your mind recently. Ooh, the conversation about AI governance is far more complex than I think many give credit. And the advice you would give a version of yourself who worked on stage at SAS Innovate, hmm, 2028? 2028, wow. Man, wear a good suit. Wear a really good suit in 2028. You know, I think, all right, I'm going to go off a tangent. Men need to wear more suits. All right, let's get men dressed again. You know, there's something that comes with, you know, being well-groomed. And so, you know, put on a jacket and go to the haberdashery, get a hat if you need to. But, yeah, come on, man, dress it up. Hopefully that'll contribute, help us contribute a little more to society in a little more meaningful way. Something's different when you put on a suit. Ah, I so much agree. I so much agree. I'm the French one, so I would agree. We have good fashion. Reggie, thank you so much for joining me today. Of course. It is my pleasure. Thanks for having me and hosting me in this beautiful space. Reggie, thank you. Every time we talk, I live with the same conviction, stated more clearly that trust isn't the price of innovation. It is the engine of it. If there is one thing to carry out of this conversation, it is this. Covenants is in the break. Done well, it is the thing that lets you move faster with your eyes open. You can follow Reggie's work at SAS and on LinkedIn, where he's one of the most thoughtful people writing on Responsible AI today. And if this gave you a single idea worth acting on before December, don't keep it to yourself. Send it to the people, the person who needs it the most. Someone on your board, your risk team, your founders. That is how an idea becomes a decision. That is Reggie Townsend. Trust isn't the cost of moving fast. It is what makes moving fast safe. I'm Sabine Van Der Linton. Keep scouting and looking forward to seeing you on the next show.