Governance isn’t a drag on competitiveness. It’s the source

This month, IMD—one of Europe’s leading business schools, where I serve as an Executive Fellow and North America Program Co-Director—released its 2026 World Competitiveness Ranking. As I pointed out in my keynote at the U.S. launch of the index at the Swiss Embassy in Washington, the results sit awkwardly with the stories that the world’s largest economies tell about their world-leading competitiveness. The truth is, neither the U.S. nor China is a true world leader in this realm—the U.S. sits in 10th place globally while China is 12th. The most competitive economies in the world are Singapore, Hong Kong, and Switzerland, three nations renowned for their prioritization of governance. IMD’s reading of the data is blunt: competitiveness now turns on institutional credibility—predictable rules, enforceable commitments, the capacity of a state to govern—more than on cost, scale, or even speed of innovation. Governance is not the price a country pays for competitiveness. It is becoming the source of that competitiveness. This cuts against a deep assumption that the less a country governs its markets and its innovators, the more competitive it will be—the view that regulation is, effectively, a tax on dynamism. In reality, the economies that take governance most seriously are also the ones that compete best. The levers of governance None of this slows innovation. Instead, the benefit governance provides is precisely that it steers innovation in the right direction. Governments hold important levers they can pull to steer how industries develop—what sort of behaviours their tax codes encourage, what their requirements are for government procurement, what information they require firms to disclose, where they place incentives. And in the case of AI, those levers can help determine whether AI implementations are aimed at replacing workers or at making them more capable. For instance, a tax code that taxes labour more heavily than investments in automation will steer businesses to automate. But that lever can also be pulled in the other direction. Leaving the lever flipped in the direction it is currently in doesn’t mean you have a low regulation economy. It means you have an economy in which you are allowing regulation to be set by historical circumstances rather than current needs. Governments have plenty of other levers to pull and buttons to push that can help them reach into less obvious areas: faced with falling literacy and numeracy, Norway is keeping generative AI out of its primary classrooms so that children still learn to read, write, and count for themselves. The outcome is a choice that can be controlled through governance. The business parallel The same logic runs through the firms that operate inside these economies. Almost every large company is now implementing some form of AI in its workflows, yet only around 5 are capturing significant value from it, according to a 2025 study by BCG. And for the most part, these models are bought, not built in house. Which means competitors can buy exactly the same technology. Whatever separates the 5 that are succeeding with AI from the rest of the marketplace cannot be the technology itself, because the technology is common property, available to all. It has to be something the AI tools cannot supply on their own. That secret ingredient is the organizational structures that decide why AI is used and how it is implemented. What determines whether an AI tool is a success or not is the set of decisions wrapped around it: where the system is allowed to operate and where it is not, what it may touch, how its output is checked and by whom, where you keep the ability to change course. Governance is that discipline. It is what turns a tool anyone can buy into a result only you could have produced. The discipline has an identifiable form—what my colleagues and I have called the CARE framework: Catastrophize the ways a deployment could fail Assess possible failure modes in depth Regulate what each system is allowed to do Prepare an exit plan that you can set in motion if the situation calls for it. The particulars will differ for every business, but the underlying point remains the same: the governance structures you put in place to guide your use of AI are a large part of the work that will distinguish you from your competitors. That is why governance is a source of advantage rather than a tax on it. Nations now compete on institutional credibility; businesses are powered by the same fuel. Three steps for tomorrow Putting in place the governance structures that will steer your AI initiatives is not something you can do overnight. But it is something you can start work on tomorrow. Here are three steps that will get you moving in the right direction. Map where AI is allowed to act. List where AI is already working in your business—not just the places you have chosen but also the shadow workstreams where team members are bringing their own AI tools to their jobs. Think through the kinds of outputs each use case involves and then decide whether an AI can make autonomous decisions in that role, develop draft ideas and texts, or should not play any part in the process. Most organizations have never drawn that line on purpose, which means it has been drawn for them, one tool at a time. Rehearse the failure before it happens. For any consequential use of AI, ask plainly how it could go wrong—a confident wrong answer, a biased pattern, a drift in scope or quality—and decide in advance what you would watch for and when you would pull it. It is far cheaper to imagine the failure now than to discover it in production. Develop metrics that capture both success and failure. Decide in advance how you will measure whether a deployment is working, and track the failures as deliberately as the wins. Most teams record the time an AI model saves them but not the extra work caused or the errors it propagates. Without good metrics, problems remain invisible and spread easily. Name who oversees outputs, and what that oversight means. Make each system, or each type of system, the responsibility of a specific person who answers for every output. Challenge them to build a comprehensive review process that is appropriate to the model and use case. Keep a route out. Before committing to a model or a vendor, write down how you would leave—what switching would cost, what you would lose, what you would need to hold in-house to maintain flexibility. Treat the absence of an exit plan as a major threat to your business integrity, even if the model you are using has always delivered in the past. Conclusion None of this is a brief for heavy regulation, and none of it asks businesses to implement fewer AI initiatives. It asks for the opposite of hesitation: a deliberate decision about how AI runs inside your company, rather than a default arrived at one tool at a time. The technology your competitors can access is the same technology you can access, so that part of the contest is already a draw. What remains open is whether you govern it better than they do—where you let AI act, who answers for what it produces, and which work stays in human hands. The economies at the top of IMD’s competitiveness ranking earned their place by treating governance as the substance of competition rather than a brake on it. For the company, the lesson is the same: the advantage doesn’t sit in the tools; it sits in the judgment you bring to implementing them.