The AI Quilter: Economic Displacement in the Age of AI Technology

Three mechanisms already in operation. None of them require intent to cause harm.

The claim that AI will reduce the economic and political power of educated women sounds either obvious or overblown depending on your starting point. What it rarely gets is specific. So let’s be specific.

There are at least three distinct mechanisms through which this plays out. They are not theoretical. Two of them are already operating at scale. The third is a structural condition that makes the other two harder to reverse.

Mechanism one: Algorithmic hiring bias

AI resume screening is not an emerging technology. It is current practice. A 2024 Gallup survey found that 93% of Fortune 500 Chief Human Resource Officers are integrating AI into hiring practices, while only about one third of employees knew their employer uses these tools. Sanford Heisler Sharp McKnight

The tools are marketed as efficient and objective. The research tells a different story. A University of Washington study tested three state-of-the-art large language models against over 550 real-world resumes and found the tools favored white-associated names 85% of the time and female-associated names only 11% of the time. University of Washington

The AI systems convert resumes into numerical representations, then measure how closely candidates match job descriptions using similarity scoring. When the algorithm gives a resume a lower score because of demographic signals, the applicant never makes it past initial screening. No human ever sees the application. The Interview Guys

The people being filtered out don’t receive a rejection. They receive silence. There is no mechanism to know it happened, no process to challenge it, and in most of the country, no legal requirement that the tools be audited for bias before deployment. California finalized AI hiring regulations in October 2025. A Colorado AI Act takes effect June 2026. New York City requires annual bias audits for automated hiring tools. There is no federal equivalent. Sanford Heisler Sharp McKnight

The tools are running. The accountability framework is not.

Mechanism two: Performance management and promotion

Getting hired is not the end of the exposure. AI bias extends through performance reviews, promotion decisions, and workplace monitoring, with algorithms trained on historical data replicating existing inequalities in career advancement opportunities. ScienceDirect

The mechanism here is straightforward. A model trained on historical promotion data learns from patterns in that data, including patterns of who got promoted, under what conditions, and by whom. If the historical record reflects a workplace that systematically undervalued women’s contributions, the model doesn’t correct for that. It learns from it. It encodes it. It applies it to the next round of decisions.

The “black box” nature of many algorithms makes it difficult for candidates and HR professionals to understand or challenge decisions. ScienceDirect This is not an incidental feature. Opacity is structural. The people affected by these decisions often have no visibility into the criteria being applied, no way to identify bias as the cause of an outcome, and no clear legal pathway to challenge a decision made by an automated system they were never told was being used.

The woman who doesn’t get the promotion doesn’t know whether it was her manager’s judgment or her algorithmic performance score. Often, neither does her manager.

Mechanism three: The structural displacement of knowledge work

The first two mechanisms operate within existing employment. The third operates at a larger scale and is harder to see until it has already happened.

The jobs being automated earliest and most aggressively are not factory jobs. They are language-based, systems-thinking, relationship-heavy roles. Writing. Analysis. Coordination. Communication. Project management. Research. The work that sits at the intersection of expertise and expression.

These are the roles disproportionately held by the educated women at the center of this conversation. And they are the roles most legible to current AI systems, because they produce outputs that can be evaluated, replicated, and iterated on by a language model.

At Davos in January 2026, Palantir CEO Alex Karp told BlackRock CEO Larry Fink that AI “will destroy humanities jobs,” adding that if you studied philosophy at an elite school, “hopefully you have some other skill, that one is going to be hard to market.” Fortune This is consistent with what he said on CNBC in March. The through-line is not a prediction. It is a description of what is already underway.

Economic leverage flows from employment leverage. When the roles that a demographic disproportionately holds shrink or disappear, the economic power that makes every other kind of leverage possible shrinks with them. Political influence, consumer power, institutional voice… these are all downstream of economic standing. That is not a controversial claim. It is how economies work.

None of this requires a coordinated effort or a singular bad actor. These three mechanisms run on their own momentum. The tools themselves start neutral. The problem enters at the moment of training, when a company feeds the model its own historical data, and that data carries every biased hiring decision, every overlooked promotion, every pattern of who got ahead and who didn’t, going back years or decades. The tool learns from that record faithfully. It has no reason to correct for it. So what gets automated isn’t a neutral process. It’s the company’s actual history, running faster and at greater scale than any human manager could.

What to pay attention to

If you are in a hiring process, you can ask whether AI screening tools are being used. Some jurisdictions now require disclosure. Most do not, but asking puts the question on record.

If you are in a workplace using AI performance management tools, you can ask how they are validated, what data they were trained on, and what the appeals process is for automated decisions. Most HR departments will not have clear answers. That is useful information.

If you are a business owner or manager making decisions about which AI tools to deploy, the question is not whether the tool is efficient. It is whether the tool has been independently audited for bias, and what you are legally and ethically liable for when it produces a discriminatory outcome. Courts are beginning to answer that question. In July 2024, a federal judge ruled that AI vendors themselves could be held liable for discrimination, not just the employers using their tools, on the grounds that drawing an artificial distinction between software decision-makers and human decision-makers would potentially gut anti-discrimination laws in the modern era. Quinn Emanuel

The tools are not going away. The question is whether the people using them, and the people affected by them, understand what they are actually doing.

That understanding is the only thing that makes an informed decision possible. Which is, in the end, the only kind of decision worth making.