Amy Bernstein (3:56)

So it's a question of speed. But you also talked about value creation and value capture. So it's also about business model. It sounds as if you're talking about, you're saying that organizations really need to rethink their entire business models. Is that accurate?

Nigel Vaz (4:14)

Absolutely. Because I think, you know, when you think about like how, you know, organizational innovation has evolved over the last so many years, right. We're really starting to see that organizations find the ability to innovate in small pockets, but they find a real challenge in how they scale these innovations across the company. And the reason that that breaks down is because largely, you know, experimentation doesn't scale unless you really reimagine how, you know, things are going to work. Right. And I think we are seeing that real big shift in the context of AI. So when you think about a lot of these proof of concepts, you know, which don't scale, they're largely because the kinds of problems they're solving are functional problems in a specific part of the organization that can deliver narrow streams of value versus the broader shift for the organization requires a bigger rethink in the context of the what and the how of, of how a business operates, which I think is most of the, the conversations we're in with a lot of companies around the world, like tackling really big meaningful problems like can you take a car that would take 18 months to redesign and, and bring that down to, to kind of 18 weeks. And if you could, what are the choices strategically does that offer you?

Amy Bernstein (5:37)

Right. And as always, it comes down to the choices. And you can tell the hard thing about strategy is what you say no to. But it sounds as if part of what you're getting at here is that for organizations that in the last decades have moved into a bunch of different businesses and that's happened across many organizations, it sounds as if what you're calling for is, is focus. And I'm wondering how you get organizations to focus strategically.

Nigel Vaz (6:08)

In my experience, the way I Think you think about focus is you've got to pick things that you can test and learn from, undoubtedly. But you have to pick those things in a way that allow you to take those learnings and make sure that those learnings actually are applicable to the broader organization. So you want to pick a problem that's not so small that it can be dismissed as irrelevant in the context of a broader transformation, but not so big that you never get out of the blocks in terms of how you actually are solving it. Right. And you want to find yourself in that sweet spot of saying this is a decent enough problem that the organization will see it as representative of how we actually solve the bigger challenge. That AI presents itself because it's forcing us to rethink, you know, so many aspects of how we engage with customers, how we drive growth, how we take cost out of the business, but at the same time it's also not so big that it just does not deliver value quickly enough. And very quickly, you know, the organization kind of moves on to the next thing.

Amy Bernstein (7:09)

So how do you advise your clients to find that sweet spot?

Nigel Vaz (7:13)

The first thing you have to get really clear headed about is what problems are you trying to solve at an organizational level and then what are the precursors to those problems that become candidates to validate that strategy? Right. Some of those questions are what questions? So what are we doing? Right. Many of those questions are how questions. I'll illustrate a kind of an example. Right. We have lots of clients who are in the midst of large scale technological transformations and they're basically looking at building new digital platforms and tools that they get out of the traditional software ecosystem. Which means that all of their business processes are baked into these software platforms that are monolithic and haven't changed or don't change frequently enough right now, rather than basically saying, hey, we're going to get rid of our ERP systems and we're going to get rid of a lot of our core technology. What you're basically saying is what are the precursors to that? So maybe we'll take a functional area which is an older application that's difficult to change, that's harder to move, and we will deploy an AI modernization effort on that area of the business. Suddenly you move from slow moving technology to an agentic agent. First orchestration, you prove the model, and now you can kind of start to say, you know what, we need to go on a broader modernization effort across our organization to replicate the learnings from here almost on an incremental basis until we are no longer bound by the business process or unconstrained by the technology that we are leading.

Amy Bernstein (8:52)

So, you know, you're talking about very new ways of thinking about organizational strategy and business strategy. And I'm wondering when you're dealing with clients, what is the thing you listen for that will most reliably kill a strategy in the age of AI? I mean, what, what is the most common error in thinking that you've come across?

Nigel Vaz (9:17)

I think probably the single biggest thing is the ability to follow a linear thought process before you get going. So this classic idea of, you know, we are going to do this, then we're going to do that, then we're going to do that, then we're going to do that, we'll review the outputs and then we'll go around the loop again. Right. And this idea of the linear baton passing, you know, functional separation of strategy. So our, we've got a corporate strategy, then we've got our finance strategy, we've got a marketing strategy, we got our product strategy, we've got our manufacturing strategy. Right. And not really focusing on thinking about how data flows across the organization and how work will get done and how these interdisciplinary tasks that create connections between sales and marketing that are historically not common but now really valuable if you could connect those data sets in the context of solving potentially a manufacturing question, not sales or marketing. Right. And being intentional around thinking about those kinds of challenges is probably the one I would highlight because it's almost like all of the success of strategic processes thus far are the very things that to some extent limit your ability to get value in terms of being intentional about how you design for an AI first world primarily around people and context and OKRs, not just technology.

Amy Bernstein (10:50)

Yeah. And what you're saying reminds me of a couple of themes that we've heard already today about the importance of trial and error. Getting away from this linear waterfall approach to strategy making, having to hammer everything to perfection before you move on to the next step. And now I'm wondering how do you know that your new strategy is working before you start getting, you know, the numbers that prove it, the, the KPIs, your OKRs, whatever they may be.

Nigel Vaz (11:23)

I think this is one of the, the biggest challenges. Right. I think this idea of like a strategic, you know, planning exercise that is separate from an executional exercise is part of that traditional model, right? I think so. Strategy today, whether it's around growth or whether it's around cost out innovation or whether it's around operational acceleration, has to come from having a strategic set of Principles and approaches, but then also from how that connects into the organization in the context of real execution, providing input back into that process. So that you don't have this idea of, look, we're going to come up with this incredible strategy. We're going to spend all this time developing a strategic hypothesis, but then, you know, we're going to sort of then deploy that validation of that hypothesis into a very linear process of measurement again. And then we'll review it at the end of next year when we do next budgeting cycle in order to iterate. Right. So much of, of this today is about measuring strategy in, in, in unit economics, not just activity. You know, thinking about the smallest possible things you can measure and using those to allow you to infer whether your strategic progress is in the direction that you want. You know, like I was using the technology examples rather than waiting for a project report at the end of a milestone. What is the cost per release? What's the cycle time per feature? What's the defect escape rate? You know, because ultimately we're in the business of helping companies transform digitally. But what that primarily means is deploying technology in order to enable either driving growth or solving cost and efficiency challenges or customer experience improvements in an organization. A lot of this comes down to how you are measuring in increments and then using that to infer or validate your hypothesis around strategic choices.

Amy Bernstein (13:26)

So you mentioned driving growth and you mentioned driving efficiency. You know, we have talked a lot about the efficiency piece of this, and I'm wondering, when you see an organization that's really using AI to drive growth, what is it doing when you're looking across your roster of clients at those that have really kind of cracked the code, what is it that they are doing differently?

Nigel Vaz (13:55)

I would say very few companies across the world would say that they've cracked the code. And we agree with that perspective. But what I can tell you about the people who are leading in the current context, right, There are a few things that they're doing that are really different. The first is recognizing that the traditional idea of software systems encapsulating a lot of the differentiation from a process perspective, is now moving to a data ecosystem of connecting different sets of data in order to start to understand how those data connections enable them to serve customers better. Whether it's in the context of, you know, improving basket sizes in a retail context by using predictive analytics on what's in that basket, and perhaps what you might be looking to create on the basis of the things you have and telling you about the few things you might have missed through to, you know, accelerating the process of the drug discovery by looking at adjacencies to the primary areas of research that the company is focused on leveraging all of the data sets from previous failed trials. Every one of these is an innovative use of connecting data into an AI first approach to creating value for end patients, citizens, customers, in a way that was just not being done historically.

Amy Bernstein (15:32)

So I want to go to some of the questions that have been coming in. Nigel, you've clearly touched a nerve with a lot of folks. One question that's gotten a lot of upvotes comes from Stacy. She says the theme throughout the summit is rethinking how we do work with that. Where do you think AI strategy should live? Is it living in IT right now? Most of the time it seems as if there needs to be a cross functional relationship with leaders, experts and individual contributors.

Nigel Vaz (16:08)

Yeah, I think that's a fantastic observation. Right. And this is where I started off right at the beginning. AI, even today, is talked about in the context of technology. And I have an analogy here to respond to Stacy's question of going back to the 90s, right? When the most valuable technology companies in the late 90s, early 2000s were companies like Cisco. Because where we were in that curve of the Internet getting established, was moving packets faster between organizations. So the whole context of the Internet conversations, and we were a company that built some of the first online banks and, and allowed you to pick a seat on an airplane. And those were not technological problems. Those were problems of business innovation. And recognizing that an airline that allows you to pick your own seat not only makes things at an airport more efficient and not only makes things that are call center more efficient. So you're not calling up and saying, Where's 32B? Because you can't see a map in front of you. It also means that it creates one of the largest revenue streams for an airline today where people are willing to pay for the privilege of picking a seat. Right. And I think this is very similar to kind of where we find ourselves with AI today. So much of the conversation is around computer. So much of the conversation today is around the technological manifestation of AI in the context of which model is better. But the reality is, whether it's compute or models, there are foundational architectural components of AI when the real conversation about AI ought to be held at a business level, because most of the value will get created on the applications, on the business processes, on the new offers we build on top of the compute and the Models. Right. And so to, I guess Daisy's question, organizations that are more successful than others are not having this conversation in the context of technology. They're having this in the context of what kind of changes are possible for us with our customers, with our employees, with our partners, and how we interact with them in the context of what is possible technologically as opposed to a technology led strategy for AI, which is fine to have at a CIO level, but that's not where the primary value unlock, I think, will come for the broader organization.

Amy Bernstein (18:28)

That makes a lot of sense. Another question that's had a lot of upvotes comes from Suzanne, who asks, when you talk about AI transformation with the human touch, what ethical red lines do you believe every organization should define fine before deploying AI at scale? How do you advise CEOs to balance the pressure for speed and cost savings with the need for responsible ethics? First, AI, especially when the short term ROI is unclear. And I'll just add, when there's so much pressure to show roi.

Nigel Vaz (19:04)

Yeah, and look, I think there's two levels of conversation here, right? I mean, one of the challenges about what makes this different than the traditional ethical discussions of the past are the fact that these ethical considerations have to be grounded in the technology. Because if you don't actually ground them in the technology, all they become is a set of ethical guidelines and principles that you put out as an organization to make yourself feel better. And what I mean by that is having a clear perspective on how are we using data that's been given to us in the context of one thing for another. What are the expectations of what kind of data we want to allow leave our organization to potentially interact with which kinds of models? What choices are we making in the context of open source models, where we can understand how models have been trained in the weighting and closed models? What are the geographic considerations in the context of sovereignty around AI and data governance, which is becoming an important consideration in the context of all of the conversations around the geopolitical landscape, you know, changing so rapidly with tariffs and other considerations, right? All of these are not just principles that can be agreed. They have to also drive a very specific set of technological decisions. So I'll give you an example of this, right? Saying we actually want to protect our customers data, but then allowing your employees to experiment on AI tools that are not in a sandbox, which is a pretty basic example, and where that data might potentially be enriching models in the public domain is an error. You think three years into this iteration of AI we would have not seen happen. But it still happens because companies aren't necessarily providing their employees tools to enable them to be the most productive that they can be. And so you are seeing somebody who is trying to help a customer in the context of a customer service problem and is finding it really hard to find the information on their own website or on the own systems they've been given, and they simply copy that question, stick it into a public AI chatbot and ask the question, sharing perhaps some of the information the customer has given them in order that they might provide that customer with a better response. But that data now of course has been, you know, has been exposed to a public domain context where you don't know exactly where and how that will, you know, propagate further. These are some basic things that I think you have to recognize aren't just now about these guidelines about seeing how we want ethical use of data, how we deal with misinformation and disinformation, how we deal with, you know, AI swap in the context of outputs, how do we deal with fakes and really helping, you know, guide in the context of marketing, perhaps or social media, what's fake and what's not, right. All of these choices have to be then embedded in system, you know, in systems and systemic ways of working in the technological approaches you choose. Because I think in this day and age, that is where the difference get made. You know, so do you have some ability to watermark or to highlight, you know, AI usage in the context of, you know, creative outputs, marketing, communications, you know, all of these things I think are where the distinction of whether you're truly living, the values that you preach around responsible AI, I think matter.

Amy Bernstein (22:42)

Yeah, that makes a lot of sense. Our next question actually is kind of adjacent to this. It's from a CEO who asks for organizations serving vulnerable communities, what safeguards are essential so AI does not not unintentionally reinforced inequities in access, voice or outcomes.

Nigel Vaz (23:02)

The most critical thing there is recognizing the data that models that you're using have been trained on in the context of services that you're providing. Right. Because we should all be clear, AI is only as good as the data that it's trained on. So if your data has all of the biases and all of the concerning components that you want eliminated from the interactions with these vulnerable communities, I think you have to start with the data sets that are being used in order to provide services. Because I think all of the things that you build on that foundation will either only compound or Potentially could be mitigated in the way that the models have been trained. And I think actually understanding how you're addressing misinformation, disinformation, information bias in the context of the data sets becomes critical. And then I think it's making sure that you have the appropriate safeguards, where you are building in reinforcement on a consistent basis around the things that you want to ensure are held true. Because I think in the context of vulnerable communities, or indeed in the context of providing equitable experiences for people, making the choice for what you want to, you know, limit is almost as important as what you want to reinforce.

Amy Bernstein (24:24)

Yeah, a lot of decision making, a lot of choices. You have to be really focused and mindful on all of that. Rich Hua, who's a founder and CEO, asks what is the most important human attribute that leaders must exhibit to successfully drive AI transformation? What social and emotional factors are leaders not thinking about Enough. This is a big HBR question.

Nigel Vaz (24:49)

We have to kind of recognize that there's something, we're at a point now where there's been a lot of conversation about AI in a kind of broad generic sense. But if you start back to this wave of generative AI, maybe three odd years ago, we started with chat and then we saw reasoning models start to evolve. And now we are in a world where we're starting to see agents, you know, very practically, I have a few on my computer now doing work while I'm having this conversation. Right. And then eventually we'll evolve to having kind of AI co workers. And I think we have to recognize this idea that the very nature of how we work as an organization is going to start to change to this interaction between us as people and the dependence and the direction and the agency that we will provide, you know, these AI tools to work on our behalf. And I think we have to start to think about this in the context of how we think about people, where you aren't necessarily just going to give them a task and they go off over a very long period of time and just kind of continue to execute that task, but you are actively able to engage with them, you know, redirect course, correct nudge and evolve. Right. And so that persistence around the memory of these agents, the interaction I think will define, I think, very strategically how we have to start to think about work in the context of organizations. Because that is very different from then how work got done just with people engaging with each other. To the second question, I think what people aren't doing is necessarily thinking as strategically about what are the Guardrails. What are the expectations if you are a chro in an organization working for a CEO like you, you know, or. Or a human resources leader, or as we call it in sapient, a people success leader, somebody whose job it is to make people successful. How do you actually start to think about making these people successful in the context of how you enable agents to interact with these people? You know, we as a business are an enterprise AI technology company in addition to, you know, a services company, right? So we think of ourselves as people and product together. And so that coexistence in organizations doesn't just exist in ours because we're in the business of providing that service that exists in the context of every business, whether you're a retailer or a telecommunications company or an airline, where your people are going to be working alongside these AI tools which have moved from being just entirely directed by people to, in some cases, you know, operating autonomously. So how do you then, in the context of that, ensure that both sides of that, you know, equation, the human and the AI coworker, as it were, are working together in a system that creates value and minimizes risk to the organization? Because I think that will be the frontier that we will find ourselves in very, very quickly.

Amy Bernstein (28:04)

So we have time for one last question, and this one got a lot of upvotes. It's from Katerina, who's a co founder. Can you give an example of effective usage of AI in strategy development or execution and what was critical to that success?

Nigel Vaz (28:20)

I'll pick an example of a strategic choice. You know, an automotive company in this instance was making on a big, you know, strategy question around how do we actually do three things at the same time. First, move quicker in a more agile environment. Second, make sure that we are being more responsive to customers, changing behaviors, and third, organizing operationally our supply chains in order to be responsive to this. Right. And the first thing I think they had to do was to strategically weight the balance of what this question was going to, you know, what this resolution was going to affect the most. And in this case, they prioritized making sure that they were being more relevant to their customers. Right. And so one of the things that they did is it shifted the process of predetermining a lot of the answers and starting to basically build what was a strategy frame around the cost, the demographic, the type of automobile that they were producing, and then almost on an iterative basis, engaging with markets to basically say, okay, if we add this camera, that allows you to reverse now that means the cost from Malaysia, which is a big audience, suddenly becomes too high, and then they're not going to participate in that model's consumption. So how do we use that information early enough that we can start to design the dash with or without a camera so that there's optional variants created then that feeds into the supply chain in terms of how they're sourcing pricing. So it's a good example of, you know, AI being used to drive large strategic decisions, you know, which then enable lots of small strategic decisions that ultimately get them to a faster design car that they're able to pivot from more quickly if they don't see as much engagement from the markets before it even hits the actual consumer, whose perspectives have been fed in, you know, through this ent entire process.

Adi Ignatius (30:38)

That was Nigel Vaz, CEO of Publicis Sapient, speaking to HBR Editor in Chief Amy Bernstein at the 2026 HBR Strategy Summit. If you found 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 and thanks to our team, Senior Producer Mary Du, Audio Product Manager Ian Fox and Senior Production Specialist Rob Eckhart. And thanks to you for listening to the HBR IdeaCast. We will be back with a new episode on Tuesday. I'm Adi Ignatius, SA.