In December 2024, just as the AI hype was ramping up, my colleagues Nicole Radziwill, Todd McClees, and I warned in the Harvard Business Review that most enterprises were poorly positioned to benefit from the technology. Our point was that AI isn’t just a technology story. It’s a human story, and people need to be set up to succeed.
Today, the evidence is in, and it is not encouraging. Study after study finds that, despite the hype, AI adoption is not having a meaningful impact on productivity. The problem isn’t the technology itself—artificial intelligence will likely be transformative—but the assumption that simply giving employees access to it will achieve much of anything.
The truth is that innovation isn’t about technology, or even ideas. As I wrote in Mapping Innovation, it is about finding novel solutions to important problems. To do that, the people who understand technologies need to learn to collaborate with people who understand real-world problems they can solve. Here’s a quick guide to scaling an idea into meaningful impact.
Start with a small majority
It’s human nature that, once we get excited about something, we want to tell everyone about it. More than that, we want to convince the skeptics to bring everyone in and get them behind the idea. That’s what happened after OpenAI made ChatGPT available for enterprise integration in March 2023. There was a mad rush by leaders to get everyone in their organization on board.
But that approach rarely works—and often backfires. Every new idea starts out weak and vulnerable. It needs to be protected. Try to move it along too fast, and you will either trigger resistance or encourage self-promoters to launch poorly thought-out projects that fail, undermining credibility and creating change fatigue.
That’s largely what happened with AI in most organizations. While most people have adopted AI for individual tasks at the lowest level of the value creation pyramid—such as writing emails or summarizing meetings—the kinds of collaboration that create greater value for customers and drive the evolution of business models remain rare. Even worse, much of the adoption has led to AI workslop that has damaged productivity and work quality.
The truth is that you don’t need to bring in everyone at once, nor should you try. You’re much better off starting with a majority of like-minded people. It might be a small, local majority of, say, three people in a room of five. As long as supporters outnumber detractors, you can move forward and gain traction in a deliberate, thoughtful way.
Do a resistance inventory
Soon after ChatGPT and other generative AI services became widely available, top-flight consultants like McKinsey, PwC, and others quickly moved to advise their clients. They issued reports touting meaningful cost reductions and AI’s role as a powerful force for transformation. What they didn’t talk much about was how and why people would resist the technology.
That’s always a mistake. Every initiative, whether it involves technology or not, will encounter resistance, and that’s why most transformations fail. Today, it is clear that the backlash against AI is growing. Data centers everywhere are facing stiff resistance, with communities pushing back against developers. Even punctuation like the em dash ( — ) has become a target.
The key to overcoming resistance is to begin before it starts. That’s why whenever we start to work with an organization, one of the first things we do is to walk them through a resistance inventory. My research has identified five categories of resistance: lack of trust, change fatigue, competing incentives, switching costs, and identity, dignity, and sense of self. We go through each, try to anticipate how they will manifest, and build strategies to overcome them.
So before you launch a new initiative, or go further with one already underway, do your own resistance inventory. Why would people not trust it? What else are they being asked to change? What incentives might compete with this initiative? What costs or inconveniences will it create? How might this conflict with how people see themselves, their skills, or their role?
You can’t anticipate everything, but you can foresee a lot just by sitting down and discussing these questions. I can’t imagine that it would have been that hard to figure out that billionaire tech bros making grandiose predictions about putting everyone out of work, and then destroying the world, might invite some skepticism.
Create a power map
We tend to think about change as a hero’s journey. There is an alternative future state, and if we do all the right things, we’ll eventually get to that place. But the true story of change is that of a strategic conflict between the status quo and that alternative future state. The status quo has sources of power keeping it in place, which always have an institutional basis.
To develop an effective transformation strategy, you need to identify the institutional stakeholders that can affect the outcome. These pillars can be divisions or functions within an organization, government agencies, regulators, legislators, advocacy groups, professional and industry associations, the media, and others. The possibilities are almost endless.
The next step is to rate each stakeholder group in terms of their levels of support and influence, to create a Power Map like the one below. This helps you see the entire transformation ecosystem and allows you to build effective strategies for each institutional group.
The Big Tech companies clearly engaged with some stakeholder groups. Certainly, their lobbying increased and they were effective in courting investors. But they neglected many important stakeholders, such as advocacy groups and others who are driving the backlash against data centers. Many unions are also deeply suspicious of the technology.
At the corporate level, many CIOs tried to implement AI as if it were just another software release and conducted little, if any, stakeholder analysis. The results were predictable. While a few enthusiastically dove in, many were resistant to using AI for anything more than low-value individual tasks.
Solving problems people care about
Every transformational effort begins with some kind of grievance: costs need to be cut, customers better served, or employees more deeply engaged. That’s what drives the need for change, no matter what technological tools you intend to use. Merely touting the power of artificial intelligence—or any technology—won’t get you far.
Far too many organizations went in search of an “AI strategy,” rather than identifying problems that people actually care about. That’s probably why so many focused on using it to reduce head count through poorly thought-out layoffs, a large portion of which have had to be reversed. They got the worst of all worlds: weakened operations and poor morale.
What’s important is that you go to where the energy is, not try to create or maintain it by yourself. That’s why we always recommend that organizations we advise to start with a keystone change, which represents a clear and tangible goal, involves multiple stakeholders, and paves the way for bigger changes down the road.
For example, when Paul O’Neill set out to turn around Alcoa in the 1980s, he started by improving workplace safety, which also paved the way to improvements in operational excellence. At Experian, when CIO Barry Libenson set out to move his company to the cloud, he started with internal APIs. In both cases, the stakeholders who were won over in achieving the keystone change also played a part in bringing about the larger vision.
As Martin Heidegger explained long ago, technology is not a creation, but a way of revealing. It unlocks new possibilities that, when channeled in the right direction, can help us solve important problems. That’s what leaders need to focus on now—not on the wonders of AI, but on how they can use new tools to solve problems that matter to people.