> ## Content Index
> Fetch the complete content index at: https://www.vitalcitynyc.org/llms.txt
> Use this file to discover other available public pages before exploring further.

# What Jobs Will AI Really Replace?
- URL: https://www.vitalcitynyc.org/ai-job-replacement-fears-reality/
- Published: 2026-11-18T16:52:00.000Z
- Updated: 2026-11-18T17:06:36.000Z
- Description: We built a desktop simulation to help keep the debate honest
- Author: Anthony Townsend
- Tags: #ai-and-the-city, Artifical Intelligence, Labor, Retraining Our Systems

The majority of service firms in the New York region, some 61%, now report using AI, the New York Fed [found in September](https://doi.org/10.59576/lse.20260901?ref=vitalcitynyc.org). That’s up from just 25% in 2024\. The share that reported laying off anyone because of AI? Only 4%. Meanwhile, Microsoft, Barclays and Kroger have joined a lengthening list of companies [citing AI](https://ny1.com/nyc/all-boroughs/ap-top-news/2026/10/23/layoffs-linked-to-changes-in-corporate-ai-spending?ref=vitalcitynyc.org) as they cut jobs. 

It’s early in this supposed economic revolution, but the questions are clear: Is AI failing to deliver automation? Or is it starting to make workers obsolete? The evidence doesn’t tell a coherent, compelling story either way — yet.

Ask a half-dozen experts and you’ll get six different answers, split roughly in two camps. One sees [a hollowing-out of white-collar work](https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/?ref=vitalcitynyc.org), a displacement of labor not seen since machines replaced mill workers. The other points to [rapid job growth at the very firms spending the most on AI today](https://alexxihe.github.io/jfe.pdf?ref=vitalcitynyc.org).

Both are, in a narrow sense, correct. Both are also wrong. We don’t yet have the evidence to forecast who will adopt AI in the future, how much AI use results in layoffs rather than better productivity, or how often that boost helps keep a worker in a job. Change your assumptions about any of these, and the answer can change by millions of jobs. In New York City, [the city comptroller’s office](https://comptroller.nyc.gov/reports/ai-and-new-york-citys-fiscal-future/?ref=vitalcitynyc.org) has modeled outcomes ranging from no job losses to an “AI Shockwave” that would cut more than 110,000 private-sector jobs in a single year. Other trends will compound this uncertainty. AI will land on an economy already being reshaped by an aging population, a weakening ability to attract immigrant workers and international students, disrupted trade and travel with Canada, and costs driven up by tariffs and the war with Iran.

We’re waiting for AI’s impacts on work and labor to come into focus. But by then, it could very well be too late.

We first ran into this problem last year during an investigation of the future of Lower Manhattan’s economy. Over the last decade, the city’s historic financial hub has experienced multiple waves of technology-driven change. Where would AI land, we wondered? Then, in March, Governor Hochul launched FutureWorks, a blue-ribbon commission charged with figuring out what AI means for the state’s workforce, which she says should be “ready to use AI to their advantage, not be victims of it.” We feared that the people answering that question were being steered by sensational reports about AI exposure, when what they needed were tools to explore and reason about the emerging evidence.

### Simulating AI’s impact on employment in New York State

That’s why our two organizations, Cornell Tech’s Urban Tech Hub and Appleseed, an economic development consulting firm, worked together to develop the [AI Job Exposure Calculator](https://ai-exposure.urbantech.info/?ref=vitalcitynyc.org). This interactive website allows anyone to map how different AI futures could play out across New York State’s workforce. We designed the underlying model and used AI coding tools to implement it.

The calculator uses simple arithmetic. First, it computes current employment in each of 845 occupations across New York State’s 10 labor markets using [IMPLAN](https://implan.com/?utm%5Fterm=implan&utm%5Fcampaign=Search%5FIMPLAN&utm%5Fsource=adwords&utm%5Fmedium=ppc&hsa%5Facc=3435734339&hsa%5Fcam=16222623499&hsa%5Fgrp=168933002628&hsa%5Fad=694993534020&hsa%5Fsrc=g&hsa%5Ftgt=kwd-329273057703&hsa%5Fkw=implan&hsa%5Fmt=e&hsa%5Fnet=adwords&hsa%5Fver=3&gad%5Fsource=1&gad%5Fcampaignid=16222623499&gbraid=0AAAAADP0vUI03piOsAd-DACZcnYa9hdXO&gclid=Cj0KCQiAwdDXBhDuARIsAFdLukme7lv1uYCPQQlcFtRuQ5jOjhrVsUnKyZLRRq3A-jsP0noPo9aRhqQaAn9dEALw%5FwcB), a commercial software tool widely used for regional economic impact analysis. Second, it isolates the share of jobs that are exposed to AI, drawing on Yale Budget Lab’s [synthesis](https://budgetlab.yale.edu/topic/artificial-intelligence?ref=vitalcitynyc.org) of six different methods for measuring the technical feasibility of AI automating work in each occupation. We then multiply this result by the share of employers who actually adopt AI to isolate the part of the workforce in occupations that are exposed to AI and whose employers actually adopt AI. A penultimate calculation further reduces to just the share affected by displacement, representing the share of automated work that actually turns into layoffs.

The end results of all these tabulations? The number of jobs at risk — where occupations are exposed to AI automating tasks, in firms that adopt AI, and where AI use leads to elimination of jobs. One final refinement adds back a few jobs to account for occupation-specific effects where research indicates AI could also help people be more productive and keep their jobs. One important caveat: what the model tells you is the size of the exposure, not the date it arrives. We simply count how many of today’s jobs would be at risk *if* a given level of AI use were reached, whether that happens in three months or three years. 

That’s the basic formula. It isn’t perfect; it’s an educated guess, akin to a weather forecast in the pre-satellite, pre-radar era. (A full discussion of our methodology and data sources can be found [here](https://ai-exposure.urbantech.info/?ref=vitalcitynyc.org#method).)

Our employment estimates are firm ground. IMPLAN’s outputs are about as good as labor statistics get. But most of the variables in our model aren’t yet well understood, and the uncertainty grows with each step. Exposure to AI has been measured more than a dozen ways, which is why the Yale Budget Lab estimates we use blend six rather than betting on one. Adoption is a moving target. Census Bureau surveys put it at roughly a third of employment and rising, and no one knows where it will top out, so we sample a range of values from 35% to 75%. Displacement is mostly bounded by historical research (we use a range of 15% to 50%). Our proxy for protection, which tries to account for instances where AI copilots help people hang onto jobs that might otherwise be automated away, is really an educated guess. We model it mostly as a coin toss.

Importantly, as we struggled to get our heads around all this uncertainty, we turned to AI ourselves. Using Claude Code, we designed a Monte Carlo simulation. This is a technique that repeatedly draws a random sample for each of its variables, many times over, to calculate how the result changes when our assumptions do. For our experiment, we spin the three global dials (representing AI adoption, AI displacement and AI protection) to create a unique scenario of AI and employment, and re-run the formula. We do it 20,000 times in a row and plot the results. The height of the curve for any given number of jobs exposed to AI roughly corresponds to the probability of that outcome. 

Remember this is a model, not the real economy. But the results can shed a lot of light on what kinds of futures might lie ahead. The first thing you notice is the fat part of the distribution, clustered around the most likely, middle-of-the-road scenario (adoption 57%, displacement 30%, protection 52%) resulting in about 628,000 jobs threatened statewide, roughly 6.3% of the labor force. Cause for concern? Certainly. It falls in the middle of the two biggest shocks of the last quarter-century, the 2008-9 recession which took roughly 335,000 jobs over an 18-month period and the COVID-19 pandemic that resulted in nearly 2 million job losses in a few weeks in March-April 2020\. But just as important, the curve shows how unlikely the worst-case scenario is. So many things have to go wrong at once — adoption at the top of its band, maximum displacement, little protection — for it to come to pass. All of this is a far cry from the alarming forecasts in recent headlines.

### What New York should do now

Armed with a clearer understanding of the range of possible risks, what should New York do now? We see three priorities where New York can lead, anticipate and act.

**Plan for mobility, not scarcity.** As [the McKinsey Global Institute has argued](https://www.mckinsey.com/mgi/our-research/workforce-in-motion-skills-and-pathways-to-future-jobs-in-the-united-states?ref=vitalcitynyc.org), long-term growth could more than offset AI-related job losses over the next decade. Even so, tens of millions of U.S. workers will at some point need to find a new job, or, harder still, a new occupation. The challenge is helping them move, delivering services for individual workers, yes, but also lower structural barriers, such as simpler state licensing, less red tape for new businesses, and portable health and retirement benefits. It also means providing more basic AI training now, through employers, colleges, libraries and expanding initiatives like the state’s new AI Jobs Institute, which aims to prepare 40,000 workers and students for AI-related jobs. 

**Shore up unemployment insurance.** Even under optimistic assumptions, several hundred thousand New Yorkers are likely to face some risk of displacement. The state has strengthened its unemployment insurance system over the past year, repaying federal pandemic-era loans, raising the level up to which wages and salaries are subject to UI taxes, and increasing the maximum weekly benefit to $869\. Nevertheless, AI-related job displacement, combined with adverse impacts of the broader trends and developments discussed above, could in the next few years drive increased spending on UI benefits, administration and services. The Department and the Division of the Budget should continue to monitor closely the financial health of the State’s UI program. 

**Ensure workers share the gains from AI.** There are already signs that the AI boom is further shifting the balance of income and wealth from labor to capital. That shift is not inevitable. Where collective bargaining is well established and AI-related job losses are plausible (as in health care, for instance), unions and industry associations should work together to establish guidelines for raising the wages of workers whose productivity is increased by AI. In non-unionized sectors with high levels of exposure to AI, such as finance, voluntary minimum wage (VMW) standards can help. Unlike the legal minimum wage, which the state sets for every employer ([$17 an hour in New York City](https://www.ny.gov/programs/minimum-wage?ref=vitalcitynyc.org) this year), VMWs are pay floors that companies adopt on their own. Since 2017, some of New York’s largest banks — including Bank of America, JPMorgan Chase, Citi and M&T have set and raised hourly floors, typically $20 to $25 by 2025; Costco and Amazon have done the same in retail.

As [Arindrajit Dube](https://www.penguinrandomhouse.com/books/708374/the-wage-standard-by-arindrajit-dube/?ref=vitalcitynyc.org) and his colleagues have shown, VMWs have helped push up wages not only for the workers directly covered, but for those slightly higher up the wage scale, and for employees of competitors. New York State should press private-sector employers to sign on to industry-wide standards, and should look for ways to extend the equity stakes that AI-company employees enjoy to the contract workers, in building maintenance, security and food service, who keep those companies running. Employees at many AI, semiconductor and fintech companies already receive a portion of their annual compensation in the form of shares, stock options, or the opportunity to buy shares at a discount, which can enable them to share in their employers’ success. New York State should seek to create similar opportunities for contract workers who contribute indirectly to AI-related companies’ success, such as building maintenance, security and food service workers. 

Perhaps the most important lesson New York can draw from its experience with AI to date is to stay nimble. AI’s effects on jobs and labor markets are evolving rapidly, and new research on those effects is appearing just as quickly. State and local officials, business and labor leaders, educational institutions and community groups can all play a role in ensuring that New York is well-positioned to reap AI’s potential economic benefits, and to ensure that those benefits are widely shared.