Strategy Summit 2026: Why AI Means Radical Change

What changes need to be made for an organization to truly succeed with their AI strategy? In this four-part special series, we’ll share conversations from the recent HBR Strategy Summit to help you get ahead. In this episode, Harvard Business School professor Tsedal Neeley shares what she’s learned about successful AI implementation and organizational transformation, from the minimum technological capabilities needed to what it takes to overcome silos to how to transform workflows and processes to add real value. HBR editor in chief Amy Bernstein facilitates, bringing in audience questions.

ALISON BEARD: I’m Alison Beard, and this is the HBR IdeaCast.

Harvard Business Review recently hosted the HBR Strategy Summit 2026, a day filled with expert advice and guidance from executives and academics alike, and we’re sharing the highlights of the event in this special IdeaCast series.

Today, you’ll hear a masterclass, an interactive lecture, from HBS professor Tsedal Neeley about how organizations can drive successful AI transformation. You’ll hear her explain the 30% rule, the minimum organizational change and baseline understanding of AI technology needed to drive real results. She’ll walk through examples from Moderna, Domino’s Pizza and Rakuten, and field audience questions facilitated by HBR Editor-in-Chief, Amy Bernstein. Enjoy the episode.

TSEDAL NEELEY: Hi, everyone. I’m delighted to be here today to talk about why AI means radical change. And what I’d like to do in the time that we have together is talk about what it takes to adopt AI at the pace that makes sense for you, for your organization, and for your industry.

AI is everywhere. People are talking about AI, AI, AI, and even agentic AI, agentic AI, agentic AI. Some of it is hype. Some of it is real. The job that we have in all of our organizations is to figure out how do we get beyond the hype and start moving at the pace that makes sense for us.

Now, first of all, I’d like to talk about the 30% rule, just to orient us on how we should think about AI and how much do we need to understand as individuals, as leaders, but also as entire workforces in order to make advancements in AI.

The 30% rule is actually a proportionality that says that we all will need a minimum technology and change capability threshold in order to contribute to a future, which has data, algorithms, and AI as part of them. And the 30% rule says you don’t need to be a programmer. You don’t need to be a data scientist. You don’t need any of those things, but you need baseline understanding like the 30% of the English language that most global employees have to master if English is not their native language.

So one of the ways to get beyond the hype is to have some baseline understanding of what AI is and what AI is not. The reality is AI is not new. It’s been around for a very, very long time. And in fact, Flagship Pioneering has captured for us four innovations or four waves of innovations in AI. The first one actually started in the 1950s. It’s called the Cybernetics Era. And in the Cybernetic Era, that’s the period where scientists at Stanford, MIT, in the military, were trying to use biology or engineering to say can we actually have machines behave in ways that have human elements? And can we actually have machines, these were very rule-based, feedback-based, behave a bit like machines? These are the early periods of robotics, actually.

And then you go to the 1980s, the 1990s. This era is called the Trained Expert Era, and this is where machines were attempting to replicate human decision-making in specific domains like medicine, like engineering, and really relying on rule-based programming and databases in order to simulate expertise in those fields. This was an effective period, not wide adoption, but we saw a big leap in AI innovation.

And then you fast-forward to the 2000s. This is a period where machine learning came to be. We started to be able to learn from the abundant data that were coming in and computing power were present and machines started to learn and to adapt. And this was kind of the early period of computer visioning, natural language processing, all of the things that got us to today.

The 2020s, generative AI with, of course, the release of ChatGPT in 2022 and increasingly agentic systems. This is where the new technology called the Transformer enables us to create new content, text, language, video, audio, big wave that has dramatically shifted the pace of how AI has been developing. But AI is not new. That’s really important to understand.

Now, I offer two definitions of AI that we need to think about. One of them doesn’t exist yet. One of them is very much in our world. The one that exists is specific AI. General AI doesn’t quite exist yet, but I will define it for you because computer scientists and philosopher of AI actually always talk about these two elements.

The one that doesn’t exist yet, general AI, you can think of this as the Terminator, where you have these systems that are human-likes. They behave like humans, they act like humans, they make decisions like humans, doesn’t exist yet. Piece of me hopes they never do. What truly does exist is what’s called specific AI or narrow AI. This is where AI performs specific tasks, much like large language models or facial recognition or voice recognition. Okay?

Now, what do we know? Many of the companies in the 2000s like Meta, used to be Facebook, Apple, Amazon, Google, Netflix, they’ve been deploying specific AI at scale in the last 15, 20 years. So we’ve actually been very much exposed to this specific AI at play in our world. So today, AI enables scale. We can serve millions and even billions of people very quickly, speed in decision-making, in operations, and scope. We can do so many more things now that we have data and we can be creative at how we get to solutions. And I’ll give us some examples about this in a little bit.

A little bit of groundwork before I do is around what I just described in terms of scale, speed and scope. What are the ways in which these are operating? Well, there are the three Ps, predictions, pattern recognitions, that’s the facial recognition example I gave you, and automation. A fourth one that’s now taking hold more and more, the fourth P is production with agents, and I’ll define that for us in a little bit.

Now, there are three vectors of value that we need to think about with AI, and this is why we say AI means change and radical change. The first one is we need to make sure that we have products that people want to use with features and functionalities that make sense for our world today. The second vector of value is network value. We want as many people using our products and services so we can not only expand those who have value from what we have to offer, but also in order to innovate. And that final vector of value is data, data, data, both internal to our organization and external data as well.

And so a flywheel of AI, as you think about what this means for us and how we need to change for this, is that the more data we have that we can harness to serve our customers, our stakeholders, or even super serve them, the better the algorithms can become or the models that we use and the better the services. So we know how to customize and personalize. And the better the services, people use our services more and that leads to more data, more data, better algorithms, better services, more usage, more data.

That’s really the flywheel. And this is where innovation truly comes through and often specific to our stakeholders, how we can super serve them. Now, the impact of AI has been quite divided. On the one hand, we know some are really leading the way with AI in one form or another, and we also know adoption has been a challenge, and I’ll talk about that as well.

So the impact of AI inside of organization, and there’s so much data today to show that people who are using AI, particularly generative AI inside of organizations, are seeing a boost in their productivity. The data, if you look at many of the studies that have been conducted at the Harvard Business School and way beyond in many other places, a one-hour task with AI used to take up to three or four hours without AI. So it’s really accelerating what people can do and how they can do them, including by using tools that typically used to be done manually.

So if we think about sales and marketing, marketing is an area that’s really been pressured and changing in the face of AI because of all the video and audio and all the images and all the content creation that’s now much faster to do. In finance or even fintech, the legal environment, HR, engineering, customer service, all of these areas, when done well, are really boosting creativity and redefining the nature of competition, which means if a firm is using AI and another is not, it becomes obvious.

On the external side, we have many examples, right? Some examples includes companies like Moderna early on. And this mindset shift that we need is captured in this quote by Stephane Bancel, the CEO of Moderna, “We’re a technology company that happens to do biology.” And what we know from Moderna early on during the COVID days, they had only 800 employees. Pfizer had 100,000 employees. Both companies were critical in the production and the delivery of the COVID vaccine in 2021 or so. You see the differences in scale.

Domino’s, we’re a technology company that happens to do pizza. Domino’s has had quite the storied run for many years in terms of its performance, because they’ve put technology and now AI at the heart of everything that they do.

Rakuten is another example of a company. I happen to sit on the board of Rakuten, and I can tell you, Rakuten, when it announced its AI strategy that’s called AInization, imagine the term, not the easiest term to say AInization, but you can imagine what that means.

And to put a much more specific point to this, the AInization strategy at Rakuten was to achieve Triple 20 business growth. And what that meant was 20% increase in marketing productivity, 20% increase in operating productivity, and 20% increase in client productivity or revenue. And this was a mandate for the entire organization to pursue and interpret in the right way, depending on the services that were the services that were functioning for each of the various businesses that Rakuten has. It has an ecosystem. And within months, the results were staggering, particularly since this was an organization mandate plus the 30% rule for everyone, and here’s some of the things that the company saw.

One is 77% decrease in marketing cost in about four months. For those who had mobile phones, Rakuten mobile phones, we saw an increase of 50% in the e-commerce side. And the adoption of AI was pretty massive. Particularly because of this mandate and a 30% rule, over 25,000 custom bots created internally within the company in a very decentralized way, empower people, equip people, and then they begin to manage their workflows. Another 800 agents as well.

Another example I can give you is what was implemented called AI Semantic Search. So if you need to make a purchase on any e-commerce site of any kind, you typically say Amazon. You typically would go on a dialogue box and you say, “Huh,” maybe it’s sneakers, “I want sneakers, women’s sneakers.” Maybe I’ll even add a size. Maybe I won’t. Maybe I’ll add a color or not. I hit Enter and then I see what comes out.

When you embed AI, specifically an AI Semantic Search like Rakuten did, you put in, “I’m going to a music festival. It’s a work event. It’s a family event. It’s a date, whatever it may be. It might be a rainy day. You know my favorite color is red. What do you recommend?” And the system would recommend a full outfit. This simple innovation has led to increase in gross merchandise sales, 6.5% here. Gross merchandise sales. If your gross merchandise sales is in the billions, imagine what this number can do.

So this is the kind of thing that leads us to say, this is about change and this is about redefining the nature of competition. I’ll give you another big example that’s taken the U.S. beauty industry by surprise and by storm. The third-largest sales platform for beauty in the United States is TikTok, and this is how it works. TikTok has over three million influencers, and this is an example of how things can play out and do play out.

This particular algorithmic AI process, an example I’m going to share with, is of this influencer. His name is Li Jiaqi. They call him The Lipstick King. And he’s an ex-L’Oreal employee, who now is an influencer with over a hundred million followers across many platforms. You see him with a product. He’s showing the product. While he’s showing the product, there are a couple of algorithms running behind the scenes, an engagement heat map. Imagine if you can do this for yourselves, and a product heat map that are determining how engaged are people and is this product going to sell?

Now, when Li Jiaqi says his catchphrase, “Oh, my God, sis, you should buy this,” people buy them in extraordinary ways. And the type of sales that you will see are things like, $3 billion in one day.

Now, think about our beauty care industries. They never saw this coming. Now, with this kind of sales, you’ve got to deliver on that, and this is why the change becomes really important. How do you deliver on this? Once you have this type of sale, you can only deliver on this if you’re structured the right way.

Number one, data integration. All systems are unified into one platform, one source of truth, allowing marketing, supply chain to be aligned with this influencer’s campaign. That’s the first thing. The second thing is the inventory management. You need a unified platform that ensures that when the influencer’s campaign is effective and there’s a spike in demand, the system automatically is updating the inventory and supply chain operations. And then consumer behavior. Consumer behavior, like I showed you, the engagement and the product heat map, the heat maps, for example, is a geometric algorithm that’s analyzing the videos, the likes, the swipes, and everything in between to determine the level of engagement. And the product engagement is actually a machine learning model that uses historical data and current data to predict whether or not this is going to sell and sell well.

And all of these put together ensures that not only is the influencer selling, but you have to fulfill these sales, and you need to be organized in a particular way to do that. And I’ll show you what that is. But first, AI agents. I’d be remiss if I didn’t take time before we transition to talk about AI agents. And what I’ll point to is it’s definitely on the rise. It’s definitely early days, and it’s definitely something that we’re going to see and hear about more and more of the next couple of years.

And essentially, what an AI agent, if you think about this, and this is Microsoft’s definition that I like a lot, many of the other Mag 7 have their own definitions, but I think this one is clear and clean. Microsoft defines AI agents as, “systems that can plan and act to complete tasks or entire workflows autonomously with key moments of human oversight.”

Microsoft released last April, and you might want to pull this, very easy to get, what they call The Year the Frontier Firm is Born. And they’ve defined aspirationally three working patterns that includes humans with assistance, humans with digital colleagues, and maybe even human-led agents. All interesting to look at as a way to think about what does this mean for us and our workflows, and to try to understand what does it mean to develop agents and these autonomously-directed workflows with human oversight.

A site that I like that can give you a sense of how this plays out is n8n.io. Check it out. Once you spend half hour on this site, it’ll clarify what all of this means. But I bring this up because there are now important reasons to figure out how we need to be organized and what’s next for us and why this is a radical change.

One is a study by one of our colleagues, Marco Iansiti, that looked at the leaders and laggards when it comes to AI. And he finds that the biggest driver of success is that with any technological architecture that you bring in, like the unified platform that I mentioned, you need to also change your processes. You need to innovate in your processes. You cannot cut and paste your old processes onto the new platform or the new approach or the new strategy that’s AI-driven in the forms that I’ve described so far.

The second thing is that AI-forward companies are organized by data, algorithms and unified platforms, not just departments. AI platforms or AI organizations look like this. This is the aspiration. You have multiple data sources, including business units that can retain and maintain their independence, but they can share their data. You can see that middle part where it says AI factory, share data with lockboxes. It’s not just you’re throwing all your data into any system. There are all these controls that you need to put in place. But if you have this AI factory, right above it, all the various units like Rakuten are sitting atop of this data platform like apps on your iPhone.

What the non-AI-forward companies, like many of our companies, look like this, siloed, the spaghetti that you see, IT projects, meetings, meetings, meetings, meetings. “Oh no, you need to go to Legal. Oh no, you need to go to this. Oh, I don’t know. No, that’s my data. No, that’s your data. No, I’m not sharing my data.” All the things, right? All organizational. But this is part of the big change that we need to get through.

Amy, I want to share one last thing and maybe you and I can talk about this. The last thing that I want to share actually is the following, existential threat, and then we can shift. So I’d like to leave you with this list, a five-point list of existential threat that can determine for you the extent to which you need to change and how you need to change. The first one is, will AI disrupt your core capabilities? The second one is, are your investor and your competitors investing in and advancing with AI? Third is client expectations. Are they changing? And boy, are they changing. We see it quite clearly in our world. Four, current tech is constraining your innovation. Think tech debt. And finally, is your culture freezing your company in outdated models?

With that, Amy, let’s talk.

AMY BERNSTEIN: Oh, my God. My chest is tight from those questions, Tsedal. Thank you.

The Lipstick King from TikTok blows my mind. But you got a bunch of questions, Tsedal. Let me share one that got an awful lot of upvotes from Kinan who asks, “How do we measure the ROI on the efficiency brought by AI?”

TSEDAL NEELEY: Ah. The ROI question is a massive question. We hear it all the time, and there are two ways to think about this. There’s ROI that you’re not going to get. For example, my dear friend and colleague that you know so well, Amy, Karim Lakhani says, “Is there ROI on WiFi?” There’s some things that we need to do where we’re not going to have direct ROI that you imagine.

On the other hand, productivity absolutely goes up with the use of AI, whether it’s software development or even creating decks and slides and analyses, et cetera. So it’s hard to say we’re going to measure this ROI.

But the thing that you need to measure is how are we innovating? What are the outcomes? We need to be very outcome-driven in this world. And even in any tech revolution that I’ve been looking at very, very closely, you need to think about what are the outcomes that we need and can we measure those outcomes? That’s where to go, the outcomes. Obsess on outcomes.

AMY BERNSTEIN: Yeah. I guess that’s always true. We should be obsessing on outcomes. Let me ask you one more question. It’s hard to choose, but let’s go with this one. From Emmanuel, “The AI hype is actually generating a lot of anxiety within our organizations. As leaders, what message can boost buy-in and engagement when adopting AI technologies?”

TSEDAL NEELEY: Every 40 years or so, there’s AI hype. It’s not new, and it could get traced back quite nicely every 40 years. We’re in that era right now. AI is going to save humanity. AI is going to destroy humanity, et cetera.

I think the best approach is number one, you have to demystify AI. 30% for everyone. And this is part of what we’re doing at the Harvard Business School. If people understand AI, it really reduces the fear and the temperature, so training, training, training.

The second thing is empirical evidence. Don’t believe the hype. Believe the proof. So we are always seeking proof, evidence, empirical, empirical, empirical.

Third, be very clear about the use cases that matter and demonstrate them and showcase them inside of your organization so that people can narrow in and focus on the use cases that are most relevant for you. But the hype is going to be there, and it’s going to be fierce.

ALISON BEARD: That was Tsedal Neeley giving a masterclass on AI as part of our recent HBR Strategy Summit. Neeley is the coauthor of the book, The Digital Mindset.

If you find this episode helpful, share it with a colleague and be sure to subscribe and rate IdeaCast in Apple Podcasts, Spotify, or wherever you listen. If you want to help leaders move the world forward, please consider subscribing to Harvard Business Review. You’ll get access to the HBR mobile app, the weekly exclusive insider newsletter, and unlimited access to HBR online. Just head to hbr.org/subscribe.

Thanks to our team, senior producer Mary Dooe, audio product manager Ian Fox, and senior production specialist Rob Eckhardt. And thanks to you for listening to the HBR IdeaCast. We’ll be back with our regular episode on Tuesday. I’m Alison Beard.