The 90-Day AI Roadmap

[00:00] Satyen Sangani (host, CEO of Alation): Welcome back to AI Radicals. My guest today has spent her career helping leaders navigate disruption, and there's no bigger disruption right now than AI. Charlene Li is a New York Times bestselling author, a keynote speaker at venues from Davos to TED, and one of the most qualified voices on leadership strategy in the world.

[00:21] Sangani: Her latest book, Winning with AI, is a 90-day blueprint for executives who need to stop treating AI as a technology project and instead start treating it as a business strategy. We get into why pilots are killing enterprise AI momentum, how to build an AI roadmap that actually connects to your business goals, and what it takes to move an organization that just doesn't want to move.

[00:41] Sangani: If you're sitting on a mandate to do something with AI but not sure where to begin, this episode is for you.

[00:48] Producer: This episode is brought to you by Alation. AI agents are only as good as the knowledge beneath them. Get it wrong and it shows up in production. Get it right with Alation at revAlation, coming to Chicago, London, and Sydney this fall, where data and AI leaders talk real measurable ROI. Save your seat at alation.com/revalation.

Why Charlene Li and Katia Walsh wrote Winning with AI

Charlene Li, co-author of Winning with AI with Dr. Katia Walsh, explains that the book started as an explainer and became a leadership and strategy book about value creation.

[01:10] Sangani: So Charlene Li, welcome to AI Radicals. You're obviously here because you're on the back of having written a new book. So let's jump right in. Tell us why you wrote the book, what motivated spending the time to do it, and what observations did you make in the process of doing it?

[01:27] Charlene Li (author, Winning with AI): My co-author Katia Walsh and I started the book because we wanted to explain what this whole phenomenon of AI is about. Then we quickly refined it to be about, how do you create value with AI? So this is not a book about the technology. It's really a leadership and strategy book for leaders to help them understand, where do you start? And that was a question that kept coming up in our conversations with them: "So what do I have to do? Where do I begin? There's so many places. It's so confusing. It's changing so quickly. What do I focus on?" And you can read all the articles and newsletters in the world, and they don't tell you how to actually use AI to create value against your strategic objective. So that's what the book is about, and it gives you 90 days to lay out a plan of how you would create value with AI, how you will win with AI.

Do you need a separate AI strategy?

Li's answer is no. The author of Winning with AI calls the belief that AI requires its own strategy the number one thing enterprises get wrong.

[02:16] Sangani: Tell us a little bit about the 90 days, and then once we get that frame, I would love to jump back into understanding how you got there. But why 90 days, and what happens in that period of time?

[02:29] Li: Well, 90 days is a quarter, and it's typically the time between when your board says you need to come up with an AI roadmap and when you have to deliver it. So that was the idea, that within those 90 days you were going to come up with an answer, like, this is what we're going to do with AI. And one of the ways we lay it out is each week is a different chapter. So there's a systematic way to go about doing this. There are many, many ways to do AI. We lay out just one of them, but we also point out ways not to do AI.

[02:59] Li: So for example, thinking that you need a separate AI strategy. No. AI is a tool that you use to support your existing business strategy, and it's the number one thing that most organizations and leaders get wrong, is they think of AI as something separate from their business strategy. It is something that is ingrained and is supportive, and you have to think about your business strategy first.

[03:21] Li: And I've seen this happen now with so many technologies, that when this bright, new, shiny object comes along, all sense and civilities around business just go out the window, and in fact you want to center everything you do around your business strategy, your mission, your purpose, your goals, your values.

[03:38] Li: And so that's what we really focus on. We lay out in all the details, how do you put out responsible AI policies, how do you figure out how you create value with AI, what your roadmap is going to look like, and the very specific building blocks to support and execute on that strategy.

How do I use AI if I don't know how to use AI?

Sangani asks why AI paralyzes leaders in a way other technologies did not. Li's answer: start with the problem you want to solve, and if you are stuck, ask AI how to use AI on it.

[03:55] Sangani: I guess it seems somewhat obvious that AI is a technology, and therefore as a technology you would sort of assume that people would say, "Oh, well, this is a thing that enables my work. This is a thing that enables what I do," whatever it is, whether you manufacture cars or fly planes or deliver food. I mean, all of these things are things that companies do. And yet it does seem like AI has uniquely taken over this position of making everybody stop and say, like, am I an AI? What do I do with this thing? And so people are not thinking of this as a business problem. They're thinking about this as a revolutionary problem. And I guess, do you feel like people are receptive to this message, that wait a second, it's a strategy problem, or is there resistance to this message, or is that changing day by day?

[04:41] Li: Well, when I first bring it up to them, they go, "Oh. Oh, I never thought of it that way." It's the first honest answer they have to that. And then when we talk about it, they go, "Yeah, I've been struggling with how I use AI." I mean, what does it mean to use AI? Because I think of it as a technology, how do I use it? I'm thinking about the prompts and the knobs and buttons I have to push. Versus say, I have this problem I want to solve. How do I use AI to solve that problem? So starting with the problem. And then what I tell them is, even if you don't know how to use AI, this is the key thing. You use AI. You ask AI how to use AI to solve your problem.

[05:19] Sangani: Yeah.

[05:19] Li: And they're like, "Oh, that's very meta." Yeah, I'm like, "Exactly." So this is a different way of thinking.

The AI translation gap between IT leaders and business leaders

Li describes a gap she sees in most enterprises: CIOs and CTOs discussing AI as technology, in front of executives who need it discussed as business.

[05:25] Sangani: Do you find that you're speaking primarily to business leaders or technology leaders as you have this conversation? Are these people who are in the IT function and in the technology function inside of the business, or are these people who are line of business leaders who are motivating strategies on the business side, or is it both?

[05:41] Li: I primarily talk to the business leaders, and there are IT leaders and technology leaders inside of there. And the reason I talk to them both in the same way is that the technology leaders need to speak in a way that the executives and business leaders understand. There is this translation gap that's happening where IT is talking about AI as a technology when they really need to be talking about it in the context of business.

[06:04] Li: So some CIOs, some CTOs are very, very good at this. The vast majority are not. And when they're given the task, go make AI work, they think of it as deploying AI into the enterprise. That is not creating value with AI. So I see wacky metrics like, we're going to get 100% adoption of AI, and I'm like, why? To what end? And there's this awkward silence that comes after that.

Should you measure AI on ROI or on business outcomes?

Li rejects return on investment as the frame for AI, on the grounds that executives do not ask for ROI on a strategy. They ask whether it delivers the outcome. It is the same tension data leaders hit when they are asked to defend data governance ROI.

[06:29] Sangani: When did you start writing the book?

[06:30] Li: We started writing in earnest in June of 2023.

[06:36] Sangani: Still before all of the run-up to what has become today. And it seems like in the news right now, there is maybe not a countertrend, but everybody in the very recent weeks has been talking a lot about ROI from AI and the reverse trend against token maxing, and there is some thought to slowing down. So in some ways, your book is coming at exactly the perfect time, because it seems like people are now craving this idea of, oh, wait a second, we actually have to think of this thing in the light of some utility or some outcome.

[07:12] Li: Right. And that was the thing that we really focus on, is how do you create value. Because it is not about return on investment. If you think about your top strategic goals, do you talk about ROI on your strategy? No, you talk about outcomes. So if you put AI in the context of your strategic goals, you would never ask about the ROI. You would say, does AI help us get to our outcomes that we want better, faster, cheaper, safer?

What does the 90-day AI blueprint actually include?

Li walks through the structure of the 90-day plan in Winning with AI. It opens with basic AI fluency for the leadership team, which is closer to a data literacy program than a technology rollout.

[07:40] Sangani: Let's talk about the 90 days. Walk us through it. What happens in the 90 days? What are my steps leaving this board meeting where my board members have told me that I need to do something? What am I doing next?

Part one: foundations, trust, and ownership

[07:53] Li: Well, we have it in three parts. The first part is just laying some foundations about getting a small team together, making sure you all know how to use AI at a very basic level. We have some exercises in there. You'd be amazed how many executives still do not understand the fundamentals of how to use AI or even the different types of AI. And then get in place some examples of responsible and ethical AI. Organizations typically have a policy. I can guarantee you that most executives, 90% of them, have not even come close to reading it and understanding what that means.

[08:25] Li: What does it mean to be responsible and, in particular, to build trust with AI? Because unless that trust is there, if there's fear and anxiety, no one's going to be able to hear what you have to say. So how do you lay those foundations for trust and responsible AI? Then think about the leadership. Who's going to be responsible for this? Somebody needs to wake up every single morning, and their job, their sole job, is to say, "How are we driving value at the enterprise level with AI?" So that's going to look very different depending on who it is in the organization.

Part two: mapping AI to the business strategy

[08:43] Li: The second part is really looking at your roadmap. How is your AI going to support your business strategy?

[09:01] Li: So assuming that everyone understands what our strategic objectives are, what are those big goals that we have? And you're assuming that everyone's agreed on that, because if you're not, that's a whole different issue. If you know that, then what are the ways you can apply it, and how do you prioritize this?

What goes on an AI roadmap instead of a list of use cases?

Li draws a hard line between a use case inventory and an AI roadmap. The roadmap commits to value delivered each quarter for 18 months, not to technology installed. A roadmap in that form can be pressure-tested against an enterprise AI checklist.

[09:19] Li: Again, this is not about starting with a bunch of use cases. Use cases are not a strategy. You have a strategy. How are you using AI to support it? And prioritize that based on the size of the impact and the speed to the impact. So we have a prioritization model for that. Then you lay out a roadmap of how you will create value, not what technology you implement, but how will you create value each quarter for the next 18 months.

[09:44] Li: So we took a lot of time to figure out what that whole center part of the 12 weeks was going to be, because it's not clear how to cut that Gordian knot. Where do you start? And we decided to start with something that organizations and their leaders know. You know your business strategy, so start with that.

How do executives get up to speed with AI?

Li reports that a minority of the executives she meets are fluent with connectors and the Model Context Protocol (MCP), and that most feel permanently behind. Her advice is to practice on real problems rather than to read more newsletters.

[10:01] Sangani: Maybe starting from the very beginning, though, in this get-to-know-AI part, it sounds like you've built a set of exercises which are just like, "Hey, let's go to ChatGPT or Claude and open up the window, and let's start actually doing some exercises and try doing your work." Do you find that some executives have still not even touched this stuff, or are there many cases where people have not tried this out?

[10:22] Li: They've tried it, and in some cases they're actively using it, so it's always a pleasant surprise. Like, okay, you're sitting here, you've got connectors going. You know what an MCP is. Awesome. This is great. They're very fluent with it. The vast majority of them don't trust it, don't use it actively, don't talk about it, feel like they are constantly behind, don't understand it.

[10:43] Li: The most common question I get asked by executives is, "How do I get up to speed with AI, and how do I maintain my knowledge? How do I keep up?" I keep framing it: do not subscribe to a dozen newsletters. And in fact, you're better off, again, starting with a problem and using AI. It's about practicing it and using it and applying it to the problems that matter to you. Because you're a leader, you know how to have very specific objectives and focus and get them done. So use AI to get your work done. And if you don't have that base ability to use AI yourself, it's very difficult to lead an organization.

[11:22] Sangani: Yeah, it absolutely is. I think the ironic question of how do I keep up with AI is to use AI, and even almost ask AI, what is it that we can do today that we couldn't do yesterday?

Should your AI policy ban AI from touching customer data?

Sangani describes an early Alation AI policy, drafted by AI, that would have barred customer records from AI entirely. The exchange is a useful test case for responsible AI in practice.

[11:34] Sangani: This question about policy is very interesting though, because we had an early funny story where we came up with an AI policy that was actually drafted by AI. And one of the things that came into the meeting when it was first presented was this idea that no data about our customers should ever interact with AI. And we were like, how is this actually going to work? We literally can't touch any records with AI? On the face of it, it sounded right, but then you said, okay, well, these policies actually don't make any sense, and on some level the policies have to evolve with the organization.

[12:11] Sangani: Do people have realistic expectations for what these policies need to be? And how do you find people in organizations think about them? Because that risk aversion is very real. I think in a world where people think about control and want to do deterministic things, they just want to lock it all down.

Goldilocks governance: how much AI governance is just right?

Li introduces two frameworks from the book. Goldilocks governance sets the amount of control, and the AI Trust Pyramid sets the order of priorities, starting from safety, security, and privacy. Both are practical entry points into AI governance.

[12:26] Li: Right. We think about governance as, we call it Goldilocks governance. It's like the children's story. Not too much, not too little, just right. And just right has to change and evolve as the world evolves, as your capabilities evolve, as your customers' expectations. And it's all about building trust. Trust in the AI, and trust in our individual and collective use of AI. Being responsible and ethical for it.

[12:45] Li: And so we have a structure called the AI Trust Pyramid. That's sort of like a Maslow's hierarchy of needs. You need to have safety and security and privacy just locked down. You want to know that it's fair and free of bias, that it's accurate, that it's responsible and accountable, and then also you're very transparent about how it's being used. And so your policies need to support that, and the definition of that will change depending on so many things.

Does AI governance slow you down or speed you up?

This is Li's central governance claim: trust is what allows an organization to move quickly, which is also the argument behind governance built into the platform rather than bolted on after the fact.

[13:24] Li: This is why we talk about it very early in the book, that when there is trust, that's when you can go fast. Good governance doesn't slow you down, it actually speeds you up, because you know what you are able to do and what you shouldn't be doing. And to your example, do we give AI the ability to look at our customer information? Yes, and it depends on which parts of that customer information. Again, privacy, safety, and security. What can you do with it? How do you know that it's accurate? All these questions that are much more nuanced than just saying, "No, don't touch it." It's unrealistic. So again, a really good AI policy is constantly being looked at, constantly being reviewed, and seeing where it's being pushed against and where you need more clarity.

[14:02] Sangani: Yeah, we believe that too. Even in something as simple as a dashboard, what you find as you develop the dashboard, and it contains all this information, and all of a sudden you're like, "Wait a second. Does everybody in the company need to see this information? Is it limited to a few people?" But you really don't even know this until you've actually developed the information and the data. And so to your point, the policies have to hit the real world, and they have to evolve.

[14:29] Producer: We're brought to you by Alation, the data intelligence platform trusted by 40% of the Fortune 100. Here's the problem they solve: your catalog helps people find data, but it doesn't actually deliver business outcomes. Alation's data products marketplace changes that. Teams can build governed, AI-ready data products in a single session using plain English, no code, no SQL. And business users can literally chat with the data to get trusted answers. Governance is baked in, not bolted on. Want to see how it works? Grab the free data sheet at alation.com/data-products.

Who should own AI in the enterprise?

Li argues the default answer, IT, is usually the wrong one, because IT has become a procurement function. Her profile for the role is closer to a change agent, which is a question of enterprise AI operating model more than org chart.

[15:10] Sangani: On this question of ownership, which I think was the third thing that you mentioned, you sort of said, like, now we have to figure out who owns this stuff. What are the ownership patterns that you're seeing in the organizations that you're working with? What does ownership mean, and how do you think about that problem?

[15:23] Li: Yeah, and we also talk about the last part of the book around governance. Again, the decision sets, all the other things around it. The number one place we see it sitting is in IT, and we're not convinced that that is the most natural place to be, because IT tends to be a procurement engine. They don't actually even build things very much anymore. And so is that the right place? If it's about procuring licenses, then that's one thing, but the actual use of it, the business strategy, we believe the person who needs to run this has to be technically oriented, but they don't have to be a technologist. Those are two different things. And then you have to have a very clear understanding of what the business strategy is going to be and understand how it can be used to support that, and then also very much a change agent, somebody who can work across silos and help people understand how, and communicate what the change is going to be.

[16:13] Li: So strategic, technical, just an amazing communicator. As you can imagine, this is a very hard role to fill.

[16:19] Sangani: Yeah, it sounds really hard.

How Moderna made its head of HR the owner of both HR and IT

Li uses Moderna as her example that there is no single correct home for AI ownership. The company moved its entire IT department under the HR leader, and Moderna's chief people and digital technology officer has since described the HR and IT merger in her own words.

[16:20] Li: And it's kind of unicorn, right? So every organization is different, and I'll give the example Moderna. Moderna had an interesting situation where they realized that their head of HR was going to be the right person to be leading this, because she understood how the workforce was going to be changed, and they were very digital already, and she was technically savvy enough to be able to manage it, so they took the entire IT department and had it report to her. So she was in charge of HR and IT. And with the idea that in the future, the way we get work done is going to be with technology and all our agents and also with people working seamlessly together. But that is their approach. That is their strategy, the way they are organized. So it's very unique to every single organization. There's no right answer.

[17:06] Sangani: Yeah, which makes a lot of sense, and I can see a lot of merits in actually having the right HR leader own transformation, because a lot of it is about getting people to think and organize differently, which makes sense, and there's a lot of enablement, to your point, that happens through HR that don't necessarily easily happen in other functions.

What if your leadership team doesn't agree on the business strategy?

Li's answer to the assumption that every executive team shares one set of strategic goals: often they do not, and siloed advocacy is what surfaces when you ask. The cross-silo problem she names is the one federated governance models exist to solve.

[17:25] Sangani: So those were the first three points. Remind me of the next stages. So the next stages were, now I'm going to sit down and I'm going to look at my business strategy. I'm going to think through how this might apply. So talk a little bit about that work. I would imagine that everybody knows what they're trying to do.

[17:41] Li: Actually, no. They may, but they may not always be in agreement about what the top strategic goals are. And the thing is that your business leaders tend to be very siloed and look at things from just their individual silos, so it's always tinged with that point of view. So they'll be advocating for the things that make sense for them, not necessarily looking at the enterprise level. So it's just going back and saying, how are we going to approach this at that enterprise level? It's not to say that the departmental and other areas aren't important, but they can handle that within the departments. We know that the cross-cutting, cross-silo areas are the hardest things for organizations to manage, and also that's where the greatest value comes from.

What are the three ways AI creates business value?

Li names the three value categories in Winning with AI: engagement, efficiency, and reinvention. Enterprises that package capability for reuse, often as governed data products, tend to find the third category easier to reach.

[18:22] Li: We go into the three different ways that you can create value with AI. The first is around engagement with customers, employees, and your various stakeholders. The second is efficiency and productivity, and it's not necessarily just cutting headcount. In fact, we think that's a very short-term and short-sighted, very scarcity-minded way of thinking about things. But it's about, if we could be more efficient, what more could we do on behalf of our customers? And then the third area is reinvention. How do you get into new business? How do you rethink your business because you have these new capabilities now? And you want to be looking across all three areas and looking at your strategic objectives to say, how can we use this perspective of engagement, efficiency, and reinvention to support our strategic objectives? So just understanding at a high level what AI can do.

[19:08] Sangani: Yeah, it feels like the efficiency piece is super well understood by the world because it's the scariest thing. It's like AI's going to take over all our jobs, and look at all these jobs in customer service that have gone away. So I think everybody kind of gets that. Talk a little about the engagement piece. What is that? What does that look like? How do people metabolize that? What does that mean?

How did Konecta use AI to grow call center headcount?

Li's counterexample to the assumption that AI guts call centers. Konecta is a large business process outsourcing (BPO) and customer experience provider, and it started from quality rather than cost.

[19:32] Li: Well, I think these two are actually very much combined. I'll give you the example of a call center. Exactly the ground zero everyone is thinking, like, this is where we're going to see the most decimation. And what we're seeing is that call centers are actually increasing their capacity and increasing their headcount. It's just the opposite.

[19:50] Sangani: So that's a broader economy stat? Like, you mean looking across all call centers, you're saying that call centers as a category in the economy are increasing headcount?

[19:59] Li: Yes. If you look at just some numbers that just came out this week where the exporting to BPOs is increasing, so people are not necessarily taking it away. What they're doing inside of those operations may differ. And then let me just give you one example. So we talked to Konecta, K-O-N-E-C-T-A, very large business processing outsourcing, call centers, customer experience, and they said from the very beginning, "We don't have a headcount problem. We have a quality problem. We don't have too many people, we don't have enough of the right people." So they used AI to increase the level of proficiency, decrease the error rates, so they removed errors by 85 to 90% with AI. They gave people co-pilots so they were able to engage with their customers more efficiently. They cut down the time because they had call summarization. And so they made them more efficient, improving their engagement.

Konecta's new business lines came out of freed capacity, not headcount cuts

Li describes what Konecta did with the capacity AI created, including a non-performing loan processing line for law firms it could not previously serve. Holding an 85 to 90% error reduction over time depends on monitoring data quality continuously rather than once.

[20:45] Li: And instead of saying, "Okay, now that we've 2x or 4x you, we're not going to cut people," we're going to say, "Well, what else can we do? We have a whole long list of things that we've never been able to do what we want to do on behalf of our customers. Now we can actually do them without increasing headcount." And on top of that, they found new businesses to get into. They were working with law firms, and they realized that they had these non-performing loans that were paying for them to do. They go, "We have this AI capability now. We can go and process those things. Would you like us to do that for you?" And the law firm is like, "Oh, yeah. Just sign us up." So whole new business lines were coming up because they had this new capability. They just announced that they were going to increase their headcount over the next two years by 5%.

[21:34] Sangani: So on some level, by doing the engagement and efficiency work, they were able to unlock this business model, business strategy capability in the third category that they otherwise may not have been thinking about or may not have been able to access.

Scarcity mindset versus abundance mindset when AI multiplies output

Li dates Konecta's decision to January 2023, immediately after ChatGPT's release, and uses it to define the two mindsets available to a leader whose people just got faster.

[21:50] Li: And what was interesting to me is when I talked to them in 2023, they began this in January 2023, right after ChatGPT came out. They could see the writing on the wall, like, "This is going to be hugely impactful on us. How are we going to approach it?" And their philosophy was, we are not going to use this to just simply cut headcount. That is just not the way to be thinking about things. And you cut headcount if you think that the world is going to stay exactly the same as it is today, and the world isn't. It is not static, and that's why it's a scarcity mindset. If instead you go, "Well, what if we use this and we enabled people," and you could 10x people, you don't say, "I'm going to cut the other nine." You go, "If I can 10x people, if I could 10x everyone, what things could be possible for us?" And that's an abundance mindset. And so this is the opportunity, I think, here.

[22:41] Sangani: You've got to really think about those problems because it really changes how you think about what opportunities you can go after. Because most of the time when you have a problem, the first thing you think about is, well, who's the responsible party for owning that problem? How do I get more people to solve that problem? And unwiring yourself from that thinking is hard to do, but also it could be quite incredible if you do it right.

Which assumptions about growth and headcount does AI force you to unlearn?

Li frames the work as unlearning, then gives the harder case: a bank whose check-reading team went from 90 people to five.

[23:04] Li: I think that's a great point, and I think it's unlearning. I think AI requires that we unlearn, and to take the assumptions that we held to be so true, for decades in some cases, and then really challenge them and say, "Is that true anymore? Do we need to add people in order to grow? Just because we can cut people, should we be? Because these are valuable people who understand our business."

[23:21] Li: Now, I talked to a bank that said, "Our AI is so good now. We used to have 90 people reading checks. Now we only need five." That's one of those cases where you just don't need the other 85 people anymore. So they said, "We could retrain and reskill and reposition about half of them," and the other half just didn't want to go through it. It just wasn't a fit for what they wanted to do. And then this is the reality.

What happens when employees don't want to be reskilled for AI?

Sangani's read on the same bank example, and on what one venture capitalist has called the Great Reorg.

[23:50] Sangani: I think that's right. I think some people don't want to be retrained, and it's in whatever role. There could be a whole set of roles. One venture capitalist wrote, I think, a paper that called it the Great Reorg, and it does feel like a little bit of the Great Reorg. It feels like some people will opt into these new roles and be totally willing to reskill and be excited to do it, and other people might be a little bit more reticent, or might take longer, or might just do different things altogether, which is okay, I think.

How do you prioritize AI investments by size and speed of value?

Li's prioritization model from Winning with AI. The double-S matrix sorts candidates by size of value and speed to value, and she manages the result as a portfolio of quick wins, momentum makers, and strategic bets. How many quick wins a team can bank depends on how quickly it can build and ship agents.

[24:18] Sangani: And so as you think about this next stage where you think about these three capabilities or these three modes of getting value from AI, then there's the business mapping and the business strategy, which I would imagine is pretty similar to any kind of strategic exercise. People can use any strategic framework, whether it's SMART, or V2MOM, or OKRs, or whatever it happens to be. Or maybe is it different? Do I think about how I strategically plan differently? Do I think about my selection differently? We talked a little bit about this mindset thing, but what are the other ways in which planning evolves and changes in an AI-filled world?

[24:54] Li: Yeah. We went around in circles with this, and we're like, if you have a strategic planning process, do it, but the biggest difference here is that you're not trying to boil the ocean. You're not trying to find the perfect solution in only a few. I really think about it as an investment portfolio. You have a portfolio of some things that we call, again, this matrix, we call it a double S matrix. The first S is size of value, and the second S is speed to value. And then based on that, you may have some things that are very quick to get value, but not that big. These are the quick wins. What are your investments going to be there? You have some that are going to be very long in value but very big, very, very big in value and take a long time. Those are your strategic bets. So where are we going to place those strategic bets? And there are some things in the medium that they're fairly good size but may take a while. Again, how do you think about these as areas of investment? And you need to have areas in all three areas, and you manage them as a portfolio.

[25:50] Li: Again, you're looking at it from a value perspective. And when you roll up the quick wins, quite a few of them, they could add up to be very equivalent to some of your momentum makers. Where, again, it may not be as fast, but are again very high in value too as well, your strategic bets. So again, treating it like a portfolio rather than sure bets is the way to go, because there are so many contingencies that you have.

Should feasibility or readiness decide what AI you build?

Li's most contrarian piece of advice. She argues feasibility studies and readiness assessments only confirm what an organization cannot do yet, and recommends a gap analysis instead, which is closer to working through an AI readiness checklist against a committed roadmap.

[26:19] Li: The one thing that we do say about this space is do not base it on feasibility. My point here is that if it's strategic, then it becomes strategic and a must-do. You wouldn't look at this and say, "Well, can we do it?" And that's how we decide how to do things. No, because you don't know how to do any of it, frankly. So feasibility is not the situation here. You may not be feasible right now. You may not have the capabilities, the technology may be there, but it should be on your roadmap. Maybe three quarters from now is when you anticipate you'll be ready to do that, but have it be on the roadmap and do the steps you need today to get to the point where you can actually implement it.

[26:58] Li: But spending any time on feasibility studies or readiness strategies and evaluation assessments, we think is just going to tell you it's not feasible and you're not ready. So how does that help you? You're better off doing a gap analysis to say, "What's our gap to actually being able to execute that, and what are the things we need to do to fill that gap?" That's a very different approach to things.

Should you start with easy AI wins or hard problems?

Sangani makes the case for hard problems on different grounds than Li's: easy problems do not change how an organization thinks, because nobody cares about them.

[27:18] Sangani: It's funny because I give people a version of that advice, which is to actually work on harder problems, not easier ones. Because often, at least the way that I come at it, is the problem with easy problems is nobody cares. And so it doesn't actually move the needle for the organization, and therefore it doesn't cause people to go change things in a way that working on a hard problem and delivering lights up the imagination and allows people to think what's different. And so I don't know if maybe I'm over fitting, but do you think that's the same thing, or how are they different? Because the feasibility version would be in the sense that, you know, that's so counterintuitive. Most people think of, "Oh, I should do the things that I can do, and I should operate within my abilities."

Why do AI pilots fail to reach production?

Li's answer starts upstream of the pilot. Work that leadership does not consider strategic gets ignored, and the word "pilot" gives everyone in the organization permission to ignore it. MIT's GenAI Divide research put the scale of it at 95% of enterprise pilots with no measurable effect on profit and loss.

[28:04] Li: Well, the thing here is that you make a good point. People don't care if it's not strategic. If it's not something that your leaders care about, if it's not in the scope of what they want, inside of the radar of what they are thinking is really important to the success of the business, they simply do not care. And so they shunt it over here: "Yeah, do whatever you want. I don't care. Don't blow things up. Just go play with your shiny objects." For them to care and put the energy and the resources of the company to do these hard things, it needs to be strategic. I have been on far too many projects that look really cool that the company leadership just really didn't care that much about. And so they always get cut.

[28:43] Li: Same thing with pilots. Don't do any more pilots. And people are like, "What do you mean? I have to test it to see if it works." I'm like, "If it's strategic, you're going to find a way to make it work." And if it's a pilot, everyone has the option to say, "Well, it's a pilot. When it's a production, when it's a project, then I'll pay attention to it. Then I'll give you the data. Then I'll give you the people. Then I'll connect you to the customers. But while it's a pilot, go off and do it yourself. I don't really need to support you and pay attention to it." But when it becomes a strategic initiative, something that is core important, and your executives are saying, "We need to do this," everyone goes, "Okay, all hands on deck. We're going to get this working and going." And that's the key to success.

Is pilot-to-production a technology gap or a leadership gap?

Li's diagnosis of why the pilot-to-production problem is everywhere in enterprise AI. Executives green-light experiments and withhold production because no AI operating system for the enterprise or roadmap exists to promote them into.

[29:26] Sangani: So this pilot question is super interesting because of the number of people that I see, the amount of marketing I'm getting from tech vendors, let alone I'm sure that many people who are in the roles that you are selling to, where it's like, move from pilot to production. I've seen this trope probably 4,000 times. It's just more than I can imagine. Talk a little bit about what's going on. Why is this pilot thing so massively prevalent in AI, and what's going on in this pilot to production thing? Why is this all happening, and why is everybody piloting everything all the time?

[30:01] Li: When I talk to vendors, they're pulling their hair out because, especially when I talk to executives, they'll go, "Can you please talk to this executive?" Because they don't know what they're doing. And so they're green-lighting these pilots because they feel they need to be doing something, but they won't green-light into production because they don't have an overarching AI roadmap that supports their strategy. They haven't put the time and attention to it. They've given it to IT. They have not really thought about what are we trying to do with this.

[30:23] Li: The way I think about it is that the business leaders have abdicated their responsibility for AI because they themselves don't understand it, and they're not prepared. They don't know. They're not equipped to be able to have a conversation, a strategic conversation with it, because they don't even understand that it is a strategic issue. So that's the gap. It's a leadership gap that we have right now, and that's the main reason why we wrote the book, is to plug that gap.

What is FOGI, the fear of getting in?

Li names the second force acting on executives. FOMO pushes them toward AI, and FOGI, fear of getting in, holds them at the edge of the water.

[30:55] Sangani: So given that currently there's a lot of technology owning AI, and to your point, business leaders have abdicated that, do you see a change in this ownership dynamic coming along? How do you see that evolving? Are there industries that lead that change more than others? What have you found in your work?

[31:13] Li: What I have found is that leaders are getting off of the sidelines. What we've been experiencing is they have this AI hesitancy. It's an AI hesitancy gap between FOMO, fear of missing out, I know I got to do something, I'm falling behind, need to move it, my board is pushing me, my people are pushing me, I need to do something, and this competing force called FOGI, fear of getting in.

[31:37] Li: And this difference of FOMO and FOGI just keeps pulling at the executive going, "I know I need to get in, but man, that water looks really murky. I don't want to jump in. I don't know what we're getting into. I don't have enough information, so I'm just going to sit here and wait for it to get clear. It feels really unsafe for me to get in. I'm staking the reputation of the company on this, and I really don't know what I'm doing, and my IT isn't helping me. No one's really telling me what I should be doing." And my advice to them is, start with what you know. Don't get frozen by the things you don't know. You know your business strategy, start with that. And that gives them a lot of reassurance, like, "Of course I know my business strategy." Now we talk about how you can use AI, and that anxiety just decreases tremendously.

What should you ask an AI vendor instead of running a pilot?

Li's script for the vendor conversation. Lead with your strategy and objectives, make the vendor do the work of mapping to them, and go straight to a production trial. Once that trial is live, evaluating AI agents against business outcomes settles whether it continues.

[32:13] Li: So what my advice to them is, when you talk to these vendors, instead of saying, "Tell me what you do," you tell them, "This is my business strategy. These are my objectives. These are the problems and the opportunities we have. How can you help me achieve these things?" And it just shortcuts the conversation, because then the vendors have to do the work. I don't want to run a pilot, I want to know whether you can help me achieve these things. If you can help me achieve these things, then we've got a deal. We'll put you in production. It'll be a three-month trial to get it up and going, and if it works, then we keep going. If it doesn't, we'll pivot. But let's go into production if you can help me with these things.

How do you lead AI change when most people don't want it?

Li cites 70% of people in the United States wanting nothing to do with AI, and treats that number as the reason leadership matters more now rather than less. Moving a number like that is a question of data culture before it is a question of tooling.

[32:58] Sangani: Give us a little bit of prediction. So you're now thinking six months out, 12 months out. Everybody has this, to your point, fear of missing out perspective. Do you think the average enterprise is very early on in their journey? Is the average enterprise well along in their journey? Where do you think people stand? And as you've done this work, how should people be feeling?

[33:21] Li: The average enterprise, the leaders know they need to do something. They're not quite sure what. They've taken it to their leaders and said, "Hey, we need to do something with AI." And they sometimes get a lot of pushback, just like, "Oh, no. We don't know how it's happening. It's very threatening to our jobs. We want nothing to do with it." And I talked to one CEO, he goes, "I really want to do something. I see there's opportunity, but my team is so not interested in moving forward, and I'm like pushing against this wall."

[33:42] Li: And the reality is 70% of people in the United States want nothing to do with AI. They think it's going to be destructive and harmful to them. And that's why leadership is even more important right now. If you truly believe AI is going to be the future, and I do believe it is, and most executives and leaders believe it is also, then you need to take a leadership moment here and push people where they don't want to go. And you have to be really strong in reinforcing that. So that's where most leaders are right now, is they're trying to figure out, how do I move this immovable force called my organization in this direction? How do I get this momentum going? How do I build a movement to get people to think more positively? And there's so much fear and anxiety around it right now. And I think the best way to do that is to have a roadmap, to explain, this is what we're going to do with AI. Let's get excited about it.

[34:36] Sangani: A lot of those roadmaps, though, require engagement from the people who are enacting them to actually build. So building that roadmap is kind of an iterative exercise, because you both have to communicate out, but you also have to listen to your team and your organization.

Will enterprise AI adoption really transform companies in the next 18 months?

Li's prediction is narrower than the headlines. Wholesale transformation stays rare, and the win looks like traction on one specific problem, which is the pattern behind most agentic data intelligence deployments that actually stick.

[34:53] Sangani: So over the next year, what do you expect to change in this enterprise adoption cycle? Everybody says, "Oh, this is the year of enterprise AI adoption," and in fact I just read somebody's predictions who thought this was going to be the year of enterprise AI adoption. And yeah, ish. Certainly the costs from all of the model providers are going up, and there are obviously use cases, coding being probably the best example, customer success or service maybe being the other example. But it doesn't feel like there's this massive groundswell of total transformation in the enterprise. Do you think that'll happen over the course of the next 12 to 18 months? How are you seeing the evolution curve?

[35:31] Li: Well, I look at it this way. The number of companies that do wholesale transformations well are very far and few between. The reality is most of them take an area, they focus on that. They take a particular problem and focus on that. And that's what I think is going to happen. We're going to see very specific problems being used, where AI is being used to solve very specific problems. And that's a great way to do it. To your point, work on hard problems. We like to say think big, start small, and then scale fast. So thinking big is strategic. Then how do you start attacking that? Well, let's take this one small problem that's a part of that and then build on that as a foundation of success.

[36:13] Li: The problem is most organizations now are not thinking big. They're thinking small, starting small, and they stay small. So I think that's going to be the transformation we see this year. It's not that suddenly AI is going to be all over the place. It is that organizations have finally found traction in some area where they're finally going, "Okay, now we're getting it. We're understanding. We're moving forward."

[36:36] Li: Transformations take a long time. Change takes a lot of heavy lifting. People do not change quickly and easily in any organization, even the fast-moving ones. So you have different people working at different speeds. So the biggest challenge that executives have is that everybody working at the same speed, so you're coordinated and moving in the same direction together, and that's just an enormous lift.

[36:59] Sangani: Yeah, for sure.

[37:00] Li: It is all about change and not about just management and operations. And most organizations and leaders are not geared towards change. They're actually managers and not leaders.

Why some CEOs don't want to lead an AI transformation

Sangani raises a survey finding on Fortune 500 CEOs who would rather avoid the transformation entirely, and asks Li what keeps her optimistic.

[37:09] Sangani: I think there was some survey, I don't remember where it came from, so I wish I had the citation. I probably ought to have a citation as a data leader before actually quoting the source, but I'll do it anyways. I think it was something to the effect of many Fortune 500 CEOs don't actually want to make this AI transformation. They actually don't want to go through the process of doing it because it seems so organizationally scary and disruptive, which tracks for me at least emotionally, because to take a 20,000, 10,000, 100,000 person organization and to try to evolve it seems really hard. Maybe inspire them for a second. What gives you the most optimism in your travels as you've seen all of this work? It feels like you have a pretty optimistic message about all of this stuff in a moment where there are lots of people who don't. What gives you the most optimism and inspiration?

What does AI look like if the goal is human flourishing?

Li's closing argument, and the building blocks behind it: mindset, skill set, tool set, and decision sets, always starting with mindset. The practical version is handing routine analysis to AI agents for data work so people spend their hours on judgment.

[38:01] Li: I'll go back to what I believe AI in the end is going to do. I do believe it's going to make us better humans. What it does is, again, there's just ways to optimize your work and be more productive and everything, but if we only just use AI to do more work, I think we would've missed the point. If instead we use AI to do all the work that we can supplement with, and then we use that extra time and capacity to develop what is uniquely human about us, our empathy, our self-reflection, our intuition, our judgment, our wisdom, if we were to elevate those things, those are the things that combine with a very strategic and deep use of AI, becomes, allows us to become truly superhuman.

[38:40] Li: So it's not just AI augmenting us, but humans augmenting AI, and you do this as an entire organization filled with superhumans, my goodness, what could you do? And if as a leader you get exhausted by the idea of making this change happen, then I really want to ask that leader, what do you think the change is toward? If it's just towards being more efficient, of course, who would get excited about that? But if you were to think about it as human flourishing, if you were to think about it as achieving things on behalf of your customers that could never have been imagined before, if you have an exercise in the world of possibilities rather than constraints, that becomes very energizing.

[39:23] Li: And so when I see organizations where leaders are energized by the idea of AI, they have that model of possibility, and that is inherent in their mindset. And it is one of the core building blocks of execution. We talk about mindset, skill set, tool set, and decision sets, but we always start with mindset. We start with the mindset of the leaders, the mindset of the culture that they create, and also shift in order to have that kind of support for that strategy, because it is otherwise just exhausting.

[39:55] Sangani: I don't know that we should leave anywhere but there. That was a phenomenal takeout for the conversation, and a lead into your book, I hope, and I hope it inspires people to go out and get it. Charlene, thank you for taking the time, and we will look forward to welcoming you back and seeing the sequel.

[40:11] Li: Okay. Thank you so much for having me.

Satyen Sangani's takeaways: AI is not the strategy, your business is

Sangani closes with the through line of the episode and the reason Li's 90-day discipline ends in production rather than in a pilot. The full framework is in Winning with AI on Amazon.

[40:15] Sangani: Charlene's core argument is deceptively simple: AI is not the strategy, your business is. That cuts against almost everything happening in enterprise AI right now. The endless pilots, the adoption metrics that go nowhere, the tech leaders handed a mandate they can't translate. Charlene traces it all back to one root cause: business leaders who haven't claimed AI as their problem to solve. Her fix is a 90-day discipline. Start with your strategic objectives, prioritize by size and speed of value, and stop calling things pilots. If it's not strategic enough to go into production, it's not strategic enough to do.

[40:45] Sangani: What stuck with me most was how she framed the end goal. Not efficiency, not headcount reduction, but instead human flourishing. AI handling more of the work so people can bring what's uniquely human: judgment, empathy, and wisdom. That's the version worth embracing. I'm Satyen Sangani, CEO of Alation. Thanks for tuning into AI Radicals. See you next time.

[41:11] Producer: Today's episode was brought to you by Alation. If you're in data or AI leadership right now, you've probably got a graveyard of pilots that never made it to production. You're not alone. MIT research shows only 5% of enterprise AI pilots from 2025 delivered real ROI. The gap isn't in capability, it's focus. Alation's agentic AI opportunity discovery guide is a strategic framework built to help you cut through the noise. It walks you through the six AI primitives, the friction-meets-value sweet spot for identifying the right processes, how to build your opportunity backlog, and how to use an impact-effort framework to prioritize what to build first. There's even a five-step launch checklist to get your first pilot off the ground. The organizations winning AI right now aren't the ones with the biggest budgets. They're the ones asking the better questions. This guide helps you ask them. It's free at alation.com/ai-guide.