Leading Through Disruption: AI Trust, the Innovator's Dilemma & Scaling AI

Victoria Mensch: Hello and welcome to Decisions at the Frontier, brought to you by Silicon Valley Executive Academy, where we help leaders tap into the Silicon Valley innovation playbook. Today's conversation is centered on a very real tension that every successful company eventually faces: the moment when the thing that made you successful starts to compete with the thing that will make you relevant — what's commonly referred to as the innovator's dilemma. My guest today has spent over a decade thinking about this tension and advising customers who are living it, across Gartner, Salesforce, Celonis and now UiPath, with a background that also includes advising on government IT policy. Christopher Radich, public sector CTO at UiPath, welcome to Decisions at the Frontier.

Christopher Radich: Thank you, Victoria. I love your mission. So excited for this one.

Victoria Mensch: Great. I just gave the audience the résumé version of how you got here. What's the version you would tell, Chris? What actually drew you to this kind of work in the first place?

Christopher Radich: You come out of school — I joined IBM at the time — just learning and learning, and then you build some confidence that you can actually be ahead of the tech curve, driving and leading some of these major technology waves. At Gartner I really had the opportunity to lean into the cloud wave at its inception and build out frameworks to help customers understand how to adopt cloud and lean into it, where there was a lot of trepidation in decision-making. That really accelerated my career. I was the adviser to the White House CIO and drove a lot of national technology policy through that role. Then I wound up in the software space, where you are in the action. You mentioned a couple of those companies — really driving transformation outcomes. A lot of it is about deploying technology and software in a safe, secure way, but a lot of transformation is about how you make confident decisions. That's where I've spent most of my career. And it's really exciting right now at UiPath, what we're doing with agentic automation. A lot of the old principles and frameworks are being refreshed, but we're looking at this problem with a fresh set of eyes: how we can help companies and government organizations adopt AI and drive outcomes.

Victoria Mensch: This is exactly the question a lot of our listeners come with, and they ask it over and over again. The thing that really keeps them up at night: how do you innovate a business that's already working? You mentioned cloud. I saw a version of this myself during my corporate career, when enterprise software companies got challenged by the shift to cloud. And you mentioned something similar happening with your customers and in your world. So what does it actually mean in practice to live through an innovator's dilemma, not just talk about it?

Christopher Radich: It's the first time I've been through it within the walls of a company. I've been more on the side of disrupting business models — for example, at Salesforce with SaaS CRM. But it's really interesting: we started in RPA, robotic process automation, 15 years ago, and have been the leader in that space. Then in 2022 ChatGPT came along with its first release — the ChatGPT moment, as it was called. A lot of organizations just thought, "All right, that's a great consumer product." I give our founder, Daniel, a lot of credit, because he saw this wave coming. Right after an earnings call, at our annual event in October 2024, he announced Act Two of UiPath. Act One was: we are the leader in rules-based automation, the leader in RPA. Now we're going to build on that and evolve into an agentic automation platform.

Christopher Radich: So, a couple of really important points. You can't wait for the market to tell you it's time to reinvent. The Act Two declaration wasn't made in 2026, when all the coding agents are out and enterprises are adopting Anthropic and OpenAI much more rapidly. Folks were just getting comfortable with frontier models. There was a lot of market research, but it wasn't data telling Daniel to do it — growth wasn't falling off a cliff, and it wasn't that no one was using RPA. You have to go with a more probabilistic — to use an AI term — instinct on this, rather than having all the clean data to finally feel confident to make the pivot.

Christopher Radich: The second part of the innovator's dilemma that I've learned through our leadership moves is that you have to evolve from a place of strength. This is not just "we're going to reinvent the company." Using the Kodak analogy: yesterday we were a physical film company, and now we're going to be something completely different. Kodak should have pivoted to digital more quickly, building on their strength. So we've built on our strength, and we're infusing our automation platform with AI capabilities to be much faster and to have more coverage of an end-to-end process. Our strength is still deterministic automation, governance, repeatability. So those are the two innovator's dilemma lessons I've learned at the leadership level: you've got to make a decision based on gut instinct as well as market research — look three years ahead, not six months ahead — and be confident in that Act Two decision. And your Act Two has to build on your core strengths as a product.

Victoria Mensch: What you're saying is that you really want to initiate changes before you're forced to. I think that's a lot of what executives are struggling with. It's one thing to have that vision and understanding, and a different thing to build consensus around it and actually move the company down that path of evolution. Is there anything you've observed at your customers or your company that helps to do that?

Christopher Radich: Repetition. As a leader, it doesn't matter what level you're at. It starts with the C-suite — in a software company it's often the founder who's going to set that Act Two vision — but as it cascades down, the C-suite is going to need lots of repetition on that message before the organization understands it's real. Then every level has to have repetition on Act Two, because it's so dramatically different. I think the default response in any company or government organization to a major change like that — a major vision or reset of direction — is "I'm going to wait this out until it becomes real." You might think the message was so clear because it was on an earnings call or at your annual event, but you have to repeat it at every opportunity you get. Repeat it with marketing alignment, with product and R&D investment, with incentives that show the organization how it's going to be measured on this new strategy. I know that sounds cliché, but it's often missed: for something to come to life, it's going to take 10, 15, 20 repetitions to your organization.

Victoria Mensch: I think that's a powerful message. Tell them, tell them again, and then tell them what you told them. But there's also the operational part you mentioned — incentives and processes aligned with the new vision. I would also assume part of that transformation is taking customers along on the journey. What does that conversation actually sound like?

Christopher Radich: It's really about translating messages. I've always been a believer that the internal message should feel no different from the customer message. To a lot of folks that sounds a bit foreign — it's almost like you have a secret code or language internally for how you talk about customers. It's a nuance. But for this to really work, you repeat the Act Two message, the conviction on Act Two and the founder's vision exactly the same way to your customers. That's the way it's landed for me personally: they feel that, and they understand this is a durable shift that's going to be successful. If you make it a product marketing thing — "here's how we're going to get you from RPA to agentic automation with a secret playbook that's only for customers" — I think you have a disconnect. It's also hard for your organization to understand two messages.

Christopher Radich: So we've really taken Act One and Act Two by leveraging the core RPA footprint. We identify the installed base — and we don't just mean the installed base of customers; we mean your RPA footprint as a customer. We go in, take a look at it, and ask how we extend it into an end-to-end, agentic, orchestrated process. Customers love that, because they've spent a lot of time and energy over the last five — some of them ten-plus — years building attended and unattended robots. When you say you can get more out of that investment and stay relevant with the trends of agentic automation, they're all ears and want to understand how. But it's all under the umbrella of the Act One, Act Two vision.

Victoria Mensch: Chris, I want to switch gears and talk more broadly about the impact of AI technology. You mentioned that AI is moving from AI that assists to AI that acts — or even the next step, merging digital AI and physical technology so it really can act. How do you help leaders build enough trust in the system to actually let it act?

Christopher Radich: This is what's holding AI back in the enterprise, and it's a nuance I don't think we talk about enough. For AI to truly be adopted in an enterprise — we're not talking consumer here — the end user has to trust that the AI is not a threat to their personal livelihood. Let's dig deeper on what that actually means. Eventually an AI system — it doesn't matter if it's a chatbot, a human-in-the-loop workflow or something more autonomous — is going to have to get through UAT, user acceptance testing. Some user is going to have to test it. And if most users feel threatened by that system — it's an efficiency play, it's going to eliminate workforce, maybe eliminate headcount a leader has — they're going to resist it.

Christopher Radich: So I always recommend a step-wise approach. This is just a little framework that I think helps. A lot of what we talk about is autonomous agents — the term agent means agency to act with full autonomy. That is not where we should start our AI adoption journey as an enterprise, because you have to work up to that level to build trust with your organization at all levels. Take middle management as an example: they'll just view it as losing all their headcount and relevance if you deploy those systems first. So I'd recommend starting with productivity chatbots. They should be on everyone's desktop, driving productivity within their role, with context grounded in that role. I see a lot of folks very comfortable with chatbots and getting into coding agents, in my organization and with most customers. You should be using this for nearly every unstructured task — at least as a thought partner.

Christopher Radich: Here's where I think the chasm is. A lot of folks think: I implement my LLM with a chatbot interface, then I jump to autonomous agents. That's where you struggle, because you'll never get through UAT. The middle tier, where I see the most success — and where we're deploying our technology for customers — is human in the loop. Human in the loop is an old automation term. It means that for any transaction to be processed or any decision to be made, a human reviews either all transactions or just the exceptions that meet certain parameters. Human in the loop demystifies things a little and takes the fear out of the autonomous agent concept. It's such a simple answer. I was just in an executive round table last week, we were talking about human in the loop, and every executive in the room wants to start building AI with this concept. We've been doing it for so long that it's second nature in the automation world. I just don't think many organizations are ready for the more autonomous back-end processing, because of all the governance, human capital and policy challenges they have there. The simple question is: are my end users going to accept this? Someone has to be the end user, and if they're going to kill it, that wastes a lot of time and money for my organization.

Victoria Mensch: Do you think there is a category of decisions that should never be handed to an AI agent, no matter how capable it becomes?

Christopher Radich: I do. That's the hard part with policy, as I talk to leaders about AI policy and grapple with it myself. We all love decision frameworks and decision trees codified into company or government policy, but it's really hard to make an ironclad policy right now. So my default would be human in the loop if a decision has a critical impact on your company — a human capital impact, a bottom-line impact, or a risk impact in a regulatory environment. That's nearly every decision a knowledge worker makes, though, which defaults back to making this an attended solution, where a human is fact-checking, reviewing and using their critical knowledge to make decisions.

Victoria Mensch: In your experience, Chris, you've probably seen some successful transitions from AI pilots to larger, company-wide AI deployment. What are the success factors you've seen?

Christopher Radich: The pilot-to-production conundrum. There are all types of studies right now about the challenges of getting from pilot to production in AI. I think you have to frame these pilots with real intent. We'll probably talk about personal failures, but whenever I've failed on any innovation work — and in AI it's all innovation work at the end of the day, which is why I love having been through previous waves like cloud — it's when there was no framing up front. The intention should be: we're going to prove technical feasibility, and we're going to prove a value hypothesis — we'll drive this much throughput or efficiency gain, maybe risk reduction or revenue impact. If we prove those things — it works in our environment, plus the value hypothesis — we're going to move it into production and roll it out. If you don't have that written contract up front with your executives and with the middle layer who's implementing it, and then communicate it down to the end users who will ultimately test it, none of these pilots will ever get into production. It's a crapshoot.

Christopher Radich: That contract up front, with entry and exit criteria for the pilot, is so important, because otherwise everyone just assumes it's a demo or not production grade. You shouldn't do work like that yourself — have your vendors do any true demo-type proofs of concept for you. But I'd even go a step further and call it a contract. This is the most formal of frameworks: if everyone doesn't agree to the contract, you shouldn't start the pilot. Make it feel really formal. I'm not talking about a 30-page document — just a couple of slides: the entry and exit criteria, and the intent that the pilot gets into a production-grade system within a certain timeline if we prove these things. Get everyone in the room, and everyone has the opportunity to raise their hand and disagree before the pilot starts.

Victoria Mensch: Speaking of this, I can imagine that contract wouldn't be an easy thing to arrive at. There are so many different parts of the organization, different stakeholders, different agendas in the room. How do you arrive at the common denominator within that contract? How do you build that consensus?

Christopher Radich: That's where I believe it takes leadership to have a vision for the outcome. I've always been a big believer — some of it is a Gartner concept or a McKinsey concept — in starting with the value hypothesis, starting with the business value up front. Don't start with the technology, or it turns into an engineering exercise that really has no exit criteria. What really drives alignment is a business objective that's been out there, that the company or government organization is driving against — then you can get to alignment much more quickly. I wouldn't expect to do that in a room by asking, "What keeps us up at night? What are our business objectives? What technology should we deploy against objective X?" That's a really hard exercise. It works when a leader frames it in a nice, clean way and says: here's a pilot initiative, intended to get into production, that aligns with objective X — which we've put a lot of resources against over the last couple of years or months — and then you have a real conversation.

Victoria Mensch: I see. There's also something you said on our call before we started recording that I want to come back to: this wave of AI adoption is so fast and so exciting that it makes you want to throw away the old playbook. But at the end of the day, you recognize that a lot of the principles are actually still the same. Can you talk about those principles? Where did you learn them, and how do you know they're still the right ones?

Christopher Radich: I've just always loved principles, because things are so noisy in any leadership role. It doesn't matter if you manage finance, IT, sales or go-to-market — you have so much coming at you as a leader. And in the AI era, why is that moving faster? Yes, the tech is moving rapidly — the model advancement. But it's mostly because people can build so much faster without technical skills. So you're getting so many more ideas coming across your desk, and principles are even more important for deciding whether to invest in an idea that someone, maybe even an individual contributor, brought to you. Here are a couple I've always used. We should write these down as leaders, because they're often timeless and they help you make great decisions.

Christopher Radich: Principle one: does this initiative — say, an AI project someone brought to your desk — align with current incentives? Your workforce is going to adopt something, and align with a process or system, when the incentives align. One thing we've learned here with the Act Two vision is that we have to align incentives. Incentives have to drive adoption of Act Two in our customer base, and they have to drive innovation internally, where we're actually using our agentic technology ourselves. When those things align, you really see things start to move.

Christopher Radich: Another one I like: as a leader, you can be very enamored with the resistors. It's almost like parenting — if a child is defiant, you want to bring them along and spend time changing their perspective. Don't pay attention to the resistors. Pay attention to those who have leaned into AI from the beginning, the folks who were ahead of the curve. There are a couple of folks in my organization I'm just so proud of — they didn't know they were this innovative. If you really pay attention, you'll find the two, three or four people who are going to drive your entire AI strategy forward. If you focus on the resistors, it's going to take a lot of energy and you're probably going to be stuck.

Christopher Radich: And the last one, which we touched on in your previous question: customers buy outcomes, not technology. Anything that feels like a very complicated technology value proposition, you need to push back on: please frame this with a business outcome, a clear KPI, a clear challenge, and show alignment with the organization's strategy. That will probably clear off a lot of the science experiments that will never make an impact. As leaders, we've never been in a position where it's so much about not only what you do but what you don't do — what you don't invest in. But you can't just kill everything, because you'll kill innovation. You have to challenge your team with those principles. Don't say, "That's just a tech science experiment, let's kill it." Say, "Come back after you've framed out the business outcome for me."

Victoria Mensch: I really love these principles, Chris. The resistors are a big part of what I'm hearing when we talk to executives who want to push the innovation agenda in their organization and are met with a "but what if it doesn't work" mentality — how to push through that and find those who say, "How can we make it work?"

Christopher Radich: Yes. And the transformation is going to spread through evidence. The early adopters are going to get you those wins. I've always loved Geoffrey Moore's Crossing the Chasm — great book. Eventually you get to the mainstream, and then the late adopters come along when it's a standard process and there's no alternative. So it really is the Crossing the Chasm approach to that principle. And I'm also enamored with some of these innovators, who are going to be your future leaders.

Victoria Mensch: So these are the principles that will help you through any technology adoption. What is genuinely different this time, Chris? What in the old playbook doesn't cover this new reality?

Christopher Radich: The new reality is that you've never been able to build as a functional individual contributor. You had to have tech skills. Even with low-code platforms — if you recall, cloud brought low-code SaaS and PaaS — you still needed a technology background. So the hardest thing to grapple with as a leader is that you're going to have so many more projects coming to your plate from all angles. I think some of the fear is driving folks to dig in on the weekends and figure out how to use these platforms and coding agents, and everybody's building agents on their own. You're going to have to make tough decisions eventually. You can't just allow a proliferation of innovation; you have to set it within your vision. You're going to have to pick some of these projects, and pick some of these emerging AI leaders in your organization, because a thousand flowers can't bloom. We've never had a wave like this, where everyone had access to the technology and there really was no barrier to building on it.

Victoria Mensch: Chris, these conversations on Decisions at the Frontier are really about decision-making, and often the most useful parts are about something that did not go as planned. Is there a decision — yours or one you watched closely — where the outcome surprised you, in a good way or a bad way?

Christopher Radich: You have to learn through failure. All growth comes through failure, as a leader and in life. There are plenty of failures I could share that have shaped these principles and my thinking, but here's a recent one. One of the concepts for bringing innovation into an organization is the hackathon. Do them with your customers; it's also very popular to do them internally, whether you're a consumer product company or a tech company. We've had many hackathons, and you see incredible collaboration and innovation. You develop this pride: wow, my organization has such great talent. But then none of the ideas make it into production. You reflect on that: it was an amazing exercise and everyone's energized, but if you do that ten times and not a single idea makes it into production, eventually the organization is going to ask, why would I bring my best stuff to this forum?

Christopher Radich: So the lesson is not that the hackathon concept is bad. If you're going to do a hackathon, set it up front: the top one or two innovation ideas or proofs of concept are going to go into production. They're going to be highlighted at the company all-hands, and they're going to be embedded within our broader AI strategy. So it's not just a random one-off. If we had done that, I think we'd have had an entirely different outcome. Because it's such a widespread technique right now and a lot of companies are doing it, I'd recommend framing it that way to really drive momentum.

Victoria Mensch: Thank you for sharing that experience, Chris. Another question, since you have so much experience with technology companies — both the more established ones and the ones in a more disruptive role: what does the high-tech industry in general get right about innovation, and what does it consistently get wrong?

Christopher Radich: Software companies, whether AI-native or transforming toward AI, do a really good job of driving constant R&D aligned with market trends. If you're a company that's not in software, or a government organization, it's worth learning how they run their P&L and continuously push R&D in line with market trends in their software category. It's what drove our Act Two product roadmap, for example. Some companies do it better than others, but software companies, US-based or international, do this really well. What do we not do well? That's harder to answer. I think software and tech companies often have such intensity around innovation that it's hard to govern it and constrain it within a more regimented strategy. With financial services providers and some other companies, you really know where they're headed with their platform investments, because they have to be so regimented in how they lay it out. Tech companies have to move so quickly, evolve and adapt, that it's not as clean.

Victoria Mensch: Now, to wrap this up: if you could leave the leaders in our audience with one principle for making better decisions when the business you're running and the business you're building are pulling in different directions, Chris, what would that be?

Christopher Radich: Listen to your customers. Listen to your customers. You know in your gut whether an investment is right or wrong, and it's always driven by your customers. Even if it's an employee decision, we should think about the level of enthusiasm we get from working with our customers to solve even some of those challenges.

Victoria Mensch: Christopher, thank you. What you're describing from the transformation point of view, from the innovator's dilemma point of view, is really helping other people think their way through. I really loved your points on listening to customers, that customers buy outcomes, not technology, on leaning on the champions and ambassadors of the technology rather than paying attention to resistors, and on aligning incentives. That's a very honest place to speak from, and it's very helpful for our listeners. Thank you so much for participating in this conversation and sharing your experience.

Christopher Radich: I really enjoyed it, Victoria. Thank you for having me on.

Victoria Mensch: Thank you. And for our listeners, if this conversation was useful, please follow our page and catch our next episode. Thank you so much.