What If AI Works Too Well
Summary: Most recently, the conversation around AI risk has focused on what happens if humans lose control. But what if the more immediate problem is what happens when AI delivers exactly the productivity we’ve been promised? Alan Chaffee considers what extraordinary AI productivity could mean for jobs, career development, and an economy in which businesses can increasingly produce more with fewer people.
There has been a palpable change in the conversation around AI recently. Some of the people closest to developing the technology are publicly raising questions about whether increasingly powerful AI systems could eventually operate beyond our control. The warnings range from calls for greater safeguards to predictions of consequences that sound like science fiction: AI becoming sufficiently capable and autonomous that humans can no longer stop it.
While I don’t discredit those fears, I don't think that's the AI risk we should be spending our time worrying about. I believe we will solve the technical problem of controlling AI. Engineers will build safeguards, redundancies, and, ultimately, some version of a kill switch. Maybe I am overly optimistic, but human beings have a good track record of recognizing existential threats and throwing enormous resources at solving them.
What I am less convinced we know how to solve is the economic problem created if AI works extraordinarily well. If it succeeds beyond our expectations.
The Problem With Extraordinary Productivity
AI has made people more productive. That is not theoretical anymore. We see it across almost every function of a business, from software development, customer service, marketing, research, operations, and finance.
The productivity gains have been unprecedented for Turning Point. Our people are spending less time gathering information, building repetitive analyses, and performing tedious work that Claude can do in minutes. They now get to spend more time diving into analysis and using the information to help clients navigate their challenges, all while getting those insights into clients’ hands faster.
For us, that additional capacity means our people can spend more time on higher-value work. But follow that productivity curve far enough and a troublesome question emerges. What happens when producing more no longer requires employing more people?
Imagine a company discovers it can increase output 30% without adding staff. That's good management. Revenue can grow faster than overhead, and margins improve. Shareholders benefit. And customers should benefit too, from better products, lower prices, or faster service.
If technology allows a competitor to produce the same product with fewer people, ignoring that technology would make your company less competitive. But an economy cannot be understood by looking at one company's income statement.
Where Do the Jobs Go?
Technology has been eliminating jobs for hundreds of years. From agricultural machinery to computers, with each disruption, new categories of employment have emerged. Maybe AI will follow the same pattern, but we shouldn’t assume history guarantees the outcome.
The International Labour Organization estimates that one in four workers globally is now in an occupation with some exposure to generative AI.1 Businesses are beginning to think differently about how much labor future growth requires. Stanford's 2026 AI Index reports that roughly a third of surveyed organizations expect AI to reduce their workforce during the coming year.2
AI isn’t eliminating all the jobs, but does the next decade of economic growth require the same amount of human labor as the last decade did?
What Happens to the First Rung of the Ladder?
Research and first drafts. Reconciliations, basic financial analysis, and building models. Summarizing information and preparing reports. Those are exactly the tasks AI is getting very good at performing.
From a productivity standpoint, automating that work makes perfect sense. But the value of those tasks goes beyond the end product.
I learned to make difficult financial judgments by first doing work that wasn't particularly difficult. I built models before I understood which assumptions mattered and prepared analyses before I was the person responsible for explaining what they meant.
If AI takes over the work traditionally assigned to people at the beginning of their careers, how do they develop the judgment we will expect from them later? The spreadsheet can tell me how many hours we save by using AI. It can't tell me what happens five years from now if nobody learned what the automated process was teaching them.
We need to think carefully about what we automate and what might be lost when we do. Productivity today can't come at the expense of developing the people we'll need tomorrow.
The Customer on the Other Side of the Equation
We also need to think beyond our own walls. Businesses employ people. People earn income. They spend that income at businesses. That cycle sits underneath an enormous portion of our economy.
Now imagine AI continues improving and businesses become extraordinarily productive, delivering more goods and services with fewer people. Maybe AI creates entirely new industries and occupations that we can't currently imagine. History gives us plenty of reasons to believe that could happen.
But there are less comfortable possibilities, too. If AI creates enormous wealth for businesses and their owners while a meaningful share of workers sees their earning power decline, we could end up with a more productive economy but fewer people able to participate in its success.
The Rational Decision Can Still Create an Irrational Outcome
None of this changes the opportunity AI presents for businesses. But as we pursue the productivity it makes possible, we need to think beyond the immediate gain.
No individual CEO is responsible for solving the future of the labor market, but they are responsible for building a competitive, healthy company, and AI will be increasingly part of doing that. The challenge is that thousands of companies are making these same decisions at the same time. What makes sense for each business individually could add up to an economic outcome none of us intended.
The Problem After the Problem
Maybe the researchers warning about AI are right, and someday we will face a machine powerful enough that turning it off becomes humanity's most important problem. In the meantime, AI doesn't need to turn against us to fundamentally change the world we live in.
I am optimistic about the enormous value that could be created. But optimism shouldn't keep us from thinking about the sum of all the parts. We may discover that controlling the machine is easier than managing the economy it creates.
---
Sources:
1. International Labour Organization, Generative AI and Jobs: A 2025 Update (Geneva: International Labour Organization, May 20, 2025), https://www.ilo.org/publications/generative-ai-and-jobs-2025-update.
2. Stanford Institute for Human-Centered Artificial Intelligence, AI Index Report 2026: Economy (Stanford University, 2026), https://hai.stanford.edu/ai-index/2026-ai-index-report/economy.
