Do you worry that you’re not getting enough ROI from AI? If so, you’re in good company. Although corporate AI adoption is growing, with one recent report showing 88 percent of organizations are now using AI in at least one function, many businesses are realizing only moderate gains or—in a small proportion of cases—no measurable value at all. For leaders, the stakes are real. Three out of four CEOs polled worldwide in 2025 said they feared they could lose their jobs within two years if that picture doesn’t change.
Of course, a tier of companies are seeing substantive benefits from AI. A look at the finer detail offers a clue to what they’re getting right. Research shows that organizations that fundamentally redesign workflows in light of AI’s capabilities are more likely to enjoy its full potential. That makes sense, as it mirrors how previous general-purpose technologies have entered the economy. Electricity came into its own within industry, for example, not as a way of electrically illuminating traditional factories, but by enabling whole new approaches to manufacturing altogether.
The move from treating AI as a bolt-on technology—something to be layered onto existing processes—to using it as a key to unlocking novel ways of working has a name: AI-native. However, for the companies not born in the AI era, embracing that ethos entails a dramatic shift in mindset, often requiring businesses to think about things they don’t even know they need to be thinking about.
Here, three leading practitioners offer some key insights for enterprise leaders going on that journey. You can find out much more about what they have to say on these topics— and more—at our five-part digital event on how to adapt, compete, and win in the AI-native era of business.
Get Comfortable With Radical Uncertainty
Benedict Evans is one of the world’s most respected technology analysts and helps leaders understand what matters most in a rapidly changing digital environment.
“With generative AI you should presume the same scale of change that we had with the internet and mobile,” says Benedict Evans. This suggests that the most successful companies of the AI era won’t merely bring new efficiencies to established business practices, but will invent entirely new kinds of AI-native products, services, and revenue strategies—just as the web gave us social media and search giants. Some of these new AI businesses may pull existing categories inside out, in the manner of Uber or Airbnb.
The challenge for leaders is that identifying these lucrative new opportunities is inherently difficult, since it’s hard to imagine things that have never been done before. The good news is that previous platform shifts have provided us with some mental models that help ensure disciplined thinking.
The first is to ask what happens when something expensive becomes significantly cheaper or free. When a new general-purpose technology dramatically reduces the cost of something, that’s what tends to unlock the more profound disruption. “If you wanted to run an express train from London to Edinburgh in 1750, it didn’t matter how much money you had. You could buy 20,000 horses and put them on the front, but it still wouldn’t work. And then steam engines come along and you could do it,” says Evans. “So what becomes cheap or free with AI? And what does that change: Does that let you compete with people you couldn’t compete with before? What new things can you now do?”
The second framework is to consider what kinds of questions computers couldn’t previously address. Take clothes shopping. “The old question is: ‘Here's a coat. Where can I buy it?’, which would work now with Google. The new question is, ‘Suggest to me 10 or 15 coats that look like this at different prices,’ which really wouldn't work before generative AI. And the next question might be something more like, ‘Look at my Instagram and suggest some coats that would freshen up my look and match what I already like.’”
Evans’ final tip is not to dismiss counterintuitive ideas out of hand. “You can't look at something and say, ‘That doesn't work.’ You have to adopt the classic venture capital framing: ‘Could it work?’” He notes that most things we do on our phones—even the idea of owning a mobile phone—would not have been a given 25 years ago. “The profoundly important new things always kind of look stupid.”
The twist to all of this is that we don't yet know the physical constraints of what will be possible with AI. From a capability perspective it’s a “moving target”, in a way that previous technologies, like mobile, were not. So this is an era in which strategic enquiry will need to keep evolving in response to developments. “Presume radical uncertainty—that everything changes, but you don’t really know how,” says Evans. “You need to be continually curious.”
AI-Native Businesses Require AI-Native IT Strategies
Michael Bradshaw is global practice leader for applications, data and AI and former CIO at Kyndryl, a leading provider of mission-critical enterprise technology services. The company helps global enterprises design, build, and operate the technical and human systems to scale AI capabilities and drive business value.
These days, a company’s technology strategy is its business strategy. Tech defines how the enterprise operates and the outcomes it’s able to deliver. But a fundamental friction has emerged: While agentic AI presents radical new opportunities, most enterprise IT was built in a different era.
Time wise, organizations designed their IT to help optimize individual business functions; as a result, their data and applications are often siloed across the enterprise. AI-native success demands a different approach, as AI’s full potential lies in its ability to cut across functions and deliver enterprise value at scale. “It’s more about changing the way the business executes holistically, from customer engagement all the way through to delivery and using that as a differentiator to drive a new dynamic in the market.”
As Michael Bradshaw sees it, achieving that can be difficult, thanks to the complexity that has entered many organizations’ IT estates over several decades. As business needs have evolved, corporations haven’t so much replaced older systems as layered new ones on top, creating what Bradshaw describes as “scaffolds”— rigid, tightly coupled environments that are hard to update.
Bradshaw’s advice is that ripping it all out and starting again isn’t necessary. Instead, organizations can work across their estates systematically, figuring out what to keep, replace or repurpose as they go. To become AI-native, enterprises need to channel this effort towards creating unified data platforms that run “horizontally” across the business. They also need to institute modular, composable architectures that can adapt as technology advances rather than relying on siloed, vendor-bound systems.
Agentic AI itself can help enable this shift towards horizontal, more agile IT. On the backend, agents can roam across systems, rather than being locked into predefined applications, accessing data directly wherever it resides. Agents can also bring greater flexibility to the end user. Today’s work spans multiple SaaS (software as a service) applications that struggle to keep pace with changing business needs. “In the old model, you get your software customized, but by the time the developers have delivered your custom software, you’ve lost the opportunity, or the business has shifted.” Now, organizations can design workflows in real time, tailored to how work happens, unlocking greater speed and adaptability.
But transforming systems alone is not enough. For Bradshaw, the real challenge of becoming AI-native is not just technical, but cultural. “It’s not a technology problem, it’s a people problem,” he says. “You can have great technology, but organizations often struggle to get employees to change.” Yet that change is the point. Becoming AI-native is about enterprises moving beyond siloed, function-led ways of working and instead aligning around shared outcomes, with teams operating across the business. “That’s a very different way of thinking about IT.”
Leadership Is Your Most Critical Technical Input
Cassie Kozyrkov, as Google’s first chief decision scientist and the founder of the field of decision intelligence, is a pioneering thinker on AI-human collaboration and the pivotal role of leadership in unlocking the true power of enterprise AI systems.
Cassie Kozyrkov often sees business leaders making a category error. They approach AI as if it were an autonomous, intelligent system, which leads them to think that human decision-making is less relevant. “But it is much more about decisions than leaders realize,” says Kozyrkov.
By way of analogy, she references the trope of the genie in the lamp. The genie has awesome power but can misinterpret wishes and act in unexpected ways. The same is true of AI, which doesn’t inherently know what is right or wrong. Getting a desired outcome comes down to how you design prompts, policies, guardrails, and harnesses, and being intentional about reviewing outcomes at regular intervals to avoid agentic drift.
It might sound like a simple statement of fact, but she is at pains to emphasise how often companies overlook the need to be “decision-first” at both a technical and strategic level. It’s vital to know precisely what you want to achieve, with what level of risk, and how you plan to measure that before you build it. It’s no good just handing out the tools and letting people do what they want with them.
That makes leadership a “technical input” into generative AI systems, she says. Without it, projects will fail.
But knowing exactly what you want to achieve from a technology that is still developing at breakneck pace is not straightforward.
One useful tactic is to keep a wish list of things you would use AI to transform in your business if the technology was mature enough—and regularly circle back to it. “It could be anything from a percentage increase in a process to a completely different product workflow,” she says. “Whatever it is, write it down, because relatively soon it may be that AI will be able to deliver that to you. Every six months at a minimum, go back to your list and see what's possible on there that wasn’t possible before.”
The nature of AI-related decisions also places new emphasis on particular skills. Unlocking the true potential of generative AI requires thinking about it not as an isolated technology but as part of a broader architecture, both in terms of its opportunities and its risks. This dual demand means that “systems thinkers”—people who naturally intuit holistic relationships, feedback loops, and patterns—are particularly valuable. “I’m seeing a lot of leaders not realizing that within their organization they actually have a lot of untapped systems thinkers,” she says. “My favorite way to spot them is to ask, ‘Who is it that everyone tends to go to when something is going wrong? Who are the people who catch mistakes?’”
Often leaders will instead look for people who might have a specific degree or qualification to help them on their AI journey, she says. “But what background really prepares anyone for this radical, brave new AI world?”
This is just a taste of what Ben, Michael and Cassie will be covering in our live broadcast “Rethink AI: Becoming AI-Native” on 5 June. Join them and other leading experts in AI and business to hear bold perspectives, practical insights, and real-world case studies to help you become AI-native. Register for free here, and have your chance to ask these experts your burning questions.
