Summary:
- AI may disrupt the labor market much like past disruptive innovations: as the cost of intelligence falls and AI improves, it could steadily move from simple tasks into increasingly complex, valuable work currently performed by people.
- Where AI is deployed will shape its impact on jobs: using it primarily to make existing markets more efficient could reduce demand for human labor, even if that disruption unfolds gradually rather than all at once.
- Nonconsumption offers another path: using AI to make previously inaccessible services affordable and available could create entirely new markets and new jobs around the infrastructure, distribution, financing, and support those markets require.
Almost everyday there’s news about the ways artificial intelligence (AI) will impact our lives. From the doomsday scenarios to the abundance the technology will unleash, there’s no shortage of opinions, academic papers, and articles about how AI will change the world. One of the things I spend my time thinking about is how AI will impact the labor market.
Some studies and AI experts (some of whom are building the technology) assert that AI will destroy many jobs while others suggest humans typically lack imagination when it comes to how new technologies will change our lives. 150 years ago, who would have imagined that we should sit in a room, stare into a rectangular screen, move our fingers, and call it work?
Admittedly, it’s hard to imagine how our lives will change the more this technology permeates every aspect of our economies. This is where I lean on Clay Christensen’s insistence that good theories help us predict the future. Data provides information about the past. To help me make sense of what might happen, I think of disruptive innovation theory. The theory suggests that AI may not simply change the conventional labor market; it may disrupt it
Disruptive innovation describes a process by which a product or service takes root in simple applications at the bottom of the market—typically by being less expensive and more accessible—and then relentlessly moves upmarket, eventually displacing established competitors. Consider how this played out in the steel industry.
In the mid-1960s, steel mini-mills, which leveraged the electric arc furnace technology, began the process of melting down scrap metal and crafting it into new steel. This process reduced costs by 20%. Initially, mini-mills could only produce the lowest quality steel like rebar, but they steadily improved and eventually tackled increasingly higher grades of steel like structural and sheet steel. By the early 2000s, companies employing mini-mills like Nucor had leveraged this innovation into a full scale disruption of the leading integrated steel makers like Bethlehem Steel and US Steel.
The causal mechanism is an asymmetry of motivation. Mini-mills were motivated to move upmarket in search of higher profits, while integrated steel companies were motivated to retreat from the increasingly unattractive lower end. We did not know whether or not the mini-mills would be able to develop technology inexpensive enough to allow them to ultimately disrupt the integrated steel companies but we knew they were motivated to keep trying. And the integrated steel mills were motivated to ignore them.
In effect, what we can confidently assert is that the cost of intelligence will continue to dwindle.
Disruption doesn’t necessarily mean complete destruction or annihilation. It does however mean that the upstart will, over time, upend the business model of the incumbent, stealthily steal their market, and most importantly, win. This, I fear, is what is happening with labor markets.
AI is following a similar trajectory. It began by performing relatively simple tasks such as drafting basic text, summarizing information, generating images, and writing simple code, but its capabilities are improving rapidly. As the cost of intelligence declines, AI will increasingly compete for more complex and valuable work, just as mini-mills steadily moved from rebar to structural and sheet steel.
At roughly $60-65 trillion, the labor market is currently the world’s largest market. The International Labor Organization estimates that the roughly 3.6 billion people who work account for approximately 52% of global GDP. Increasingly, AI is capable of performing some of the tasks for which this enormous market currently pays human beings.
We are still at the infancy of the technology and it hasn’t been integrated into most of our workflows like the Internet, mobile phone, or computer. But adoption is happening fast and productivity will ensue as a result. As such, AI’s impact on the labor market is unlikely to be immediate and sudden. Like mini-mills, it’ll happen over time.
History suggests that technological shifts rarely eliminate industries overnight, but they do reshape where growth and value creation occur. As disruptive innovation shows, many new technologies often begin as inferior alternatives, and are dismissed by incumbents. Over time, as performance improves and cost structures shift, the center of gravity of value creation moves. It happened with transistors and vacuum tubes, personal computers and mainframes, digital photography and film, streaming and video rental stores, and cloud computing and on-premise servers. The incumbents did not disappear entirely. But they were no longer the primary engines of growth.
I’m having a hard time seeing how this doesn’t happen with labor (the incumbents) as the diffusion of AI continues to occur at a rapid pace.
Nonconsumption as the antidote to major job losses
But there’s hope.
Nonconsumption is the inability of an entity (person or organization) to purchase and use (consume) a product or service required to relieve a struggle. This inability to purchase can arise from the product’s cost, inconvenience and complexity, along with a host of other factors — none of which tend to be limitations for the rich, skilled, and powerful in society. And globally, nonconsumption is vast.
Fewer than 7% of the global population lives on more than $50 a day, or $18,250 a year. In Africa, South Asia, and Middle East and North Africa it’s 0.3%, 0.1%, and 0.8% respectively. This means that most people in the world don’t have access to basic resources like food, quality education and healthcare services, social safety nets, retirement funds, housing, and so on.
If the power of AI is unleashed to create access, it would result in new markets that effectively create new jobs for people. Consider the billions of people who cannot afford a tutor, financial advisor, lawyer, doctor, software developer, or business consultant. If AI dramatically lowers the cost of accessing some of these capabilities, it enables people who previously could not consume these services to become consumers, thereby creating new markets around them. And as these new markets form, they create demand for the distribution, sales, infrastructure, financing, support, management, and complementary services required to serve them.
However, if AI targets existing markets and makes them more efficient, it’s likely to destroy more jobs than it creates. This would lead to a disruption of the existing labor market. The disruption will not happen overnight and human labor will not entirely disappear just as integrated steelmaking did not disappear. But it may no longer be the primary engine of growth and value creation across many parts of the economy.
The question, then, is not simply whether AI will destroy jobs. It is where we choose to point this increasingly abundant and inexpensive intelligence. If we primarily deploy it against existing consumption, we should expect cost efficiencies, increasing firm profits and significant pressure on existing jobs. But if we deploy it against the vast nonconsumption that still defines much of the world, AI could help us build entirely new markets and, as a result, new jobs. The future of work may depend less on what AI can do than on which problems we choose to solve with it.
