The Economy Needs Fewer Workers and More Entrepreneurs. How Are We Equipping Them?

A friend came to me recently with a question I could not answer easily.

“My boss just used AI to build a dashboard for the accounts that I was supposed to be managing. What should I do?”

I did not have a good answer. Because the honest answer is that it was not just the dashboard. It was the signal. If your boss can replace your output with an AI tool in an afternoon, the role itself is on borrowed time. And my friend knew it. That look on his face was not confusion. It was the quiet realisation that the ground had shifted.

That moment stayed with me. Because it is not one person’s problem. It is the shape of things to come for millions of working professionals who did everything they were told to do, and are now discovering that “doing what you are told” is exactly the skill set AI replaces first.

The Question Nobody Wants to Answer

Senior leaders, investors, board members. I want you to touch your heart and answer this honestly.

If your business no longer requires 100% of your workforce, would you cut headcount to save costs, or would you find a way to keep everyone?

Most leaders want to keep their people. But wanting to is not the same as being able to. There comes a point where the money you are spending to keep roles alive is money you cannot afford to burn.

You are faced with a hard choice: let people go and remain competitive, or keep everyone, absorb the cost, and try to find new ways to generate revenue before the runway runs out.

Some companies will find that path. Most will not. And the longer you wait, the harder the choice becomes.

But if every company makes the rational choice, and the number of available jobs shrinks structurally, where do the people go?

Organisations Are Getting Structurally Leaner, and the Jobs Are Not Coming Back

My friend’s dashboard story is not an anomaly. It is the pattern.

In Singapore, DBS Group announced plans to cut around 4,000 contract and temporary staff over three years as AI increasingly takes on roles carried out by people. Not through dramatic layoffs, but through natural attrition: as contracts end, the roles will not be replaced because AI can do the work.

Klarna, the Swedish fintech, cut its workforce by roughly 40%, from about 5,500 to around 3,400 by the end of 2024. Revenue more than doubled. Average pay for remaining staff rose from US$126,000 to US$203,000. The company did not fire people in a dramatic layoff. It simply stopped replacing staff who left, and let AI absorb the work.

Anthropic, the company behind Claude, ran its entire growth marketing operation with a single person for nearly ten months. Austin Lau, who had no prior programming background, used Claude Code to build the full performance marketing stack. This is at a company approaching US$7 billion in annual revenue.

Midjourney generated roughly US$500 million in revenue in 2025 with around 100 to 160 employees. No venture capital funding. Over US$5 million in revenue per employee.

Companies no longer need large headcounts. They can operate hyper-lean, with AI agents serving as different “departments,” handling marketing, customer service, data analysis, and operations that used to require entire teams.

In previous disruptions, there was always a bounce-back. In the dot-com crash, the internet economy stumbled, but it did not change the fact that companies still needed people to get work done. The equilibrium state of the economy still assumed human labour as the primary input. Companies recovered. They rehired.

The AI transition is different. The new equilibrium state, the one the economy is moving toward, is one where AI handles a significant share of cognitive work. Unlike previous transitions, this new equilibrium does not include companies rehiring at the same scale. The work can now be done by AI. This is fundamentally different in scale and in kind.

The International Monetary Fund (IMF) estimates that roughly 60% of jobs in advanced economies are exposed to AI. About half of those may benefit through productivity gains. The other half face displacement pressure. The PwC 2024 Global AI Jobs Barometer found that sectors most exposed to AI experienced 4.8 times higher labour-productivity growth than less-exposed sectors.

More productive. Fewer people needed.

If Companies Cut and AI Replaces, Where Do People Go?

Here is the logic chain that most discussions about the “future of work” avoid stating plainly. And I think they avoid it because it is uncomfortable.

Companies adopt AI. They become more productive with fewer people. This is the rational, efficient choice. Across an entire economy, this means the total number of available jobs shrinks. Not because of a recession. Not because of a temporary market correction. Because the work itself can now be done by fewer humans.

In previous economic disruptions, the displaced could retrain and find new employment because new industries emerged that still needed human labour at scale. Factories replaced farms. Service industries replaced factories. Tech companies replaced traditional services. Each wave created new categories of jobs.

The honest question is whether AI creates enough new job categories to absorb the people it displaces. Nobody knows. But here is what we do know.

Retraining does not create jobs. Government programmes do not create jobs. Intrapreneurship creates value inside existing organisations, but it does not create new employment at scale.

The only mechanism in economics that creates net new jobs is entrepreneurship. New businesses, serving new needs, hiring new people.

That is a confronting reality. Because most of the people who will be displaced were never taught how to do this.

A Swedish register study (von Greiff, 2009) found that job displacement roughly doubled the probability of entering self-employment, from about 1.4% to roughly 2.6%. Only a small percentage of displaced workers historically became entrepreneurs.

But that finding reflects a world where starting a business required significant capital, engineering skills, and distribution networks. All three barriers have been dramatically reduced by AI. That 2.6% is a floor from a different era, not a ceiling for this one. The question is not whether more people will attempt entrepreneurship. They will, because they will have to. The question is whether we equip them for it, or whether we leave them stranded with a skill set built for a world that no longer exists.

Sam Altman’s OpenResearch guaranteed-income experiment, published as an NBER working paper in 2024 (Vivalt et al.), gave 1,000 low-income individuals US$1,000 per month for three years. The key finding: recipients worked one to two fewer hours per week and were more likely to hold out for meaningful work. They did not start businesses in significant numbers.

What does this tell us? That money alone does not make someone entrepreneurial. Entrepreneurship requires a mindset, a skill set, and a support system. And for most people, these were never part of their education. They were taught to be reliable, to specialise, to follow instructions. The entire school-to-employment pipeline was designed for the industrial age. AI is eating exactly that profile.

The Frameworks We Learned in Business School Are Outdated

For decades, entrepreneurship education has rested on a handful of canonical frameworks. Porter’s Five Forces taught us to analyse industry structure. Blue Ocean Strategy taught us to find uncontested markets. The Lean Startup taught us to validate fast and fail cheap. Peter Thiel’s “Zero to One” taught us to build creative monopolies.

These frameworks shaped how an entire generation of business leaders thinks about competition, strategy, and starting ventures. And every one of them was built on assumptions that AI is now dismantling.

Porter’s Five Forces assumed stable barriers to entry.

Michael Porter’s framework rests on the idea that industries have relatively fixed structures: capital requirements, distribution advantages, regulatory moats, and supplier relationships that take years to build or break. Strategy meant finding a position within that structure and defending it.

AI collapses barriers to entry across knowledge-work and software-based industries. A solo founder with access to AI coding tools can now build what used to require a funded engineering team. Cloud-hosted AI APIs mean you no longer need proprietary technology to compete. Several scholars are now proposing AI adoption as a “sixth force” that reconfigures the original five rather than simply feeding into them.

The implication: competitive advantage is less about occupying a defensible position and more about how fast you can learn, experiment, and adapt.

Blue Ocean Strategy assumed uncontested markets could last.

W. Chan Kim and Renee Mauborgne’s Blue Ocean Strategy is built on the premise that firms can create “uncontested market space” through value innovation and then enjoy years of profitable growth before competitors catch up.

AI shortens that timeline dramatically. The moment someone identifies a novel customer need, AI-powered analytics let competitors spot the same pattern. Generative AI tools mean a competing product can be prototyped in days, not months. Platform distribution (app stores, search, social media) means incumbents can bundle similar capabilities and ship them to massive user bases almost immediately.

Blue oceans still exist. They just close in months instead of years. The skill is no longer finding one blue ocean and defending it. It is the ability to continuously identify, enter, and move on from short-lived opportunities.

Daniel Priestley, in a recent conversation on The Diary of a CEO, put it simply: the barrier to entry is now so low that blue oceans get crowded before you finish building your dock.

The Lean Startup assumed building was slow and expensive.

Eric Ries published “The Lean Startup” in 2011. The core idea, the Build-Measure-Learn loop, was revolutionary because building a product used to take months and cost significant capital. The whole point of a Minimum Viable Product was to reduce that waste.

AI has compressed the build phase to near-zero for many product categories. AI coding assistants let non-engineers create functional web and mobile applications. Reports document non-programmers building apps in under an hour using AI tools. Design, copywriting, and even basic user research can be AI-assisted.

The Build-Measure-Learn loop has not disappeared. It has collapsed into something closer to continuous experimentation. The bottleneck is no longer “can we build this?” It is “do we have the judgment to know which experiments matter?”

Thiel’s Zero to One assumed software moats were durable.

Peter Thiel and Blake Masters argued in “Zero to One” (2014) that startups should seek creative monopolies through unique technology, network effects, and strong distribution. Implicitly, the book assumes that building such a product requires scarce technical talent and significant time, making imitation difficult.

AI commoditises software creation. AI-powered coding tools like Cursor have reached over $1 billion in annualised revenue with fewer than 300 employees, precisely because they make it trivial for anyone to build what used to require a full engineering team. If the tool to build software is itself a commodity, then the software you build with it is doubly commoditised.

The monopoly opportunity still exists, but it has shifted. It is no longer about a unique product. It is about owning a category, a workflow, a community, or a distribution channel. Features can be copied in weeks. Trust, relationships, and ecosystems cannot.

The critical takeaway: these four frameworks are not wrong. They are outdated. They were designed for a world where building was expensive, barriers were high, and competitive advantages lasted years. In a world where AI makes building cheap, barriers low, and advantages temporary, we need a different playbook.

The Honest Tensions

Three things complicate this argument, and it would be dishonest to ignore them.

First, not everyone can or should be an entrepreneur. Risk tolerance is not evenly distributed. Cultural expectations in many Asian societies prioritise stability, family obligation, and established employment over entrepreneurial risk. Research on entrepreneurial intentions across six ASEAN economies shows that social norms and family expectations significantly shape who even considers starting a business. In Singapore, where government and large-firm employment is culturally valued, “starting a business” carries different social weight than it does in Silicon Valley.

Second, AI could get expensive. The current “cheap AI” environment is subsidised. OpenAI, Google, and Anthropic are spending billions to acquire users. Current pricing is below cost. This is the classic platform strategy: subsidise adoption, build dependency, adjust pricing later. Data centre infrastructure has a three- to four-year lifespan before the chips become obsolete, unlike roads or fibre that last decades. Nobody knows whether AI costs will continue falling or whether the current pricing is a temporary acquisition phase. Prudent entrepreneurs build businesses that survive even if AI costs rise.

Third, the distribution of AI’s gains is uneven. I wrote about this in a previous article: AI multiplies output, but it does not multiply the money in the economy. If every company uses AI to produce more with fewer people, but the total economic demand does not grow proportionally, the efficiency gains flow upward to margins and ownership, not downward to the people absorbing the productivity load. Competition is the mechanism that distributes value more broadly. Without new businesses entering markets, incumbents have no pressure to share gains. Entrepreneurship is not just a career path. It is the economic force that prevents AI’s benefits from concentrating at the top.

So What Do We Actually Need?

If entrepreneurship is the mechanism that creates new jobs, but most people are not equipped for it, then the gap is not “more AI tools.” The gap is what I call Domain Expertise multiplied by Entrepreneurship, multiplied by AI.

Most AI adoption today is still at what I have described in a previous article as Level 1: people using ChatGPT, Copilot, or Gemini to write emails and summarise documents. Useful, but not transformative. The gap between chatbot-level AI use and actually building AI-powered products, services, and workflows is where the real opportunity sits.

What the displaced, the ambitious, and the forward-thinking actually need is a combination of three things.

Domain expertise. Your lived experience, your industry knowledge, what you have seen and done that no AI can replicate. Daniel Priestley calls these “personal playbooks.” They are the starting point for every business that solves a real problem.

Entrepreneurial thinking. The ability to spot unmet needs, validate ideas quickly, build minimum viable offerings, and iterate. Whether you are starting your own company or proposing a new revenue line inside your current employer, this is the skill set that creates new value.

AI as the capability multiplier. Not AI as a chatbot. AI as a tool to build, automate, distribute, and scale. Programmes like Creating Agentic Workflows teach exactly this: how to build systems that create business value, not just automate existing tasks. The person who can combine domain expertise with entrepreneurial thinking and use AI to execute at speed has a structural advantage that no amount of prompt engineering can match.

Domain expertise multiplied by entrepreneurial thinking, multiplied by AI. That is the formula.

At Overpowered, this is the thesis behind AI³ (AI to the power of 3). AI¹ is Appreciative Inquiry: a strengths-based process to surface what you already know and connect it to real market needs. AI² is Artificial Intelligence: learning to build custom tools, not just use chatbots. AI³ is Applied Intrapreneurship: using those tools to create new value, whether inside an organisation or as an independent venture.

The reason I built this programme is because I kept seeing the same pattern in conversations with talent development leaders across industries: people know AI exists. They use it to draft emails. But they cannot imagine what comes next. The ceiling is not the technology. It is the imagination.

“A lot of people know AI. But they don’t really KNOW AI. They’re attracted by the flashy things but don’t really know how it can help.”
// VP of Talent Management, top-three global logistics operator

Intrapreneurship matters here too. Not everyone will start a business. But many people can think entrepreneurially inside their current organisation, if their leaders create space for it. The constraint is not the employee’s capability. It is often the leader’s willingness to let people experiment, fail, and build something new.

If you are a leader reading this: the question is not whether your people can think entrepreneurially. It is whether you are willing to let them.

The Choice Ahead

Let me be honest about what I think will happen. Most companies will cut. They will look at the numbers, see that AI can do the work of three people, and make the rational decision. A few, the good ones, will find ways to redeploy their people, to invest in their growth, to use AI as a tool that multiplies what their team can do rather than a reason to shrink it.

But regardless of which type of company you work for, the shift is happening. People will be impacted. Not in some distant future. Now. The contract staff at DBS. The customer service teams at companies following Klarna’s model. The junior analysts, the administrative staff, the marketing coordinators whose roles are quietly being absorbed by AI tools. People like my friend, staring at a dashboard that used to be his job.

If you are a leader: the question is what kind of leader you want to be. One who cuts and moves on, or one who equips your people for what comes next.

If you are an employee: the question is whether you wait for your organisation to decide your future, or whether you start building the skills that make you valuable regardless of what any single employer does.

If you are an educator, a policymaker, or someone who designs training programmes: the question is whether you are still teaching frameworks built for the industrial age, or whether you are preparing people for an economy where entrepreneurial thinking is not optional.

They deserve more than a LinkedIn post about “upskilling” and a link to a free online course.

The old rules of entrepreneurship are outdated. Porter, Blue Ocean, Lean Startup, Zero to One. They were built for a world that no longer exists. The new rules are faster, leaner, and more uncertain. And if we are serious about preparing people for what is coming, we need to start teaching them.

Because if the economy needs fewer people inside companies, it needs more people building new ones. And those people need to know that they can.