The Jobs Automation Quietly Removed

A sourced accounting of industrial robots, self-checkout, kiosks, electronic tolling and a ten-percent autonomous-trucking scenario

By Micheal "Cupcake" Cobb

August 13, 2026

Automation does not usually arrive with a pink slip attached to every machine.

It arrives as eight self-checkouts watched by one employee. It arrives as two ordering kiosks where two people once took orders. It arrives when a retiring worker is never replaced, when an open position quietly disappears, or when a company produces more with fewer people. That is why the public argument over automation is so slippery: the machines can be counted, but the jobs they eliminate, suppress or reshape are rarely recorded in one place.

There is no federal database that totals every American job lost to robots, kiosks, automated banking, electronic tolling and software-controlled machinery. The responsible approach is to keep three categories separate: documented employment changes, calculated job-equivalent scenarios and future projections. This analysis does exactly that.

Bottom line: The corrected equipment-based model produces approximately 2.94 million net job/FTE-equivalents since the 1990s, or 3.08 million gross. A broader and disputed manufacturing-productivity method produces approximately 5.82 million. Neither number is a census of individual people fired on a particular day.

The robot number that keeps getting misused

The International Federation of Robotics reported a preliminary 38,000 industrial-robot installations in the United States during 2025, up 11 percent from 2024. Automotive accounted for 13,500 installations, while food-industry adoption rose 30 percent. IFR separately reported that 393,700 industrial robots were operating in U.S. factories at the end of 2024. Those are different measurements: annual installations are a flow; operational stock is the number believed to remain in service. (IFR, 2026; IFR, 2025)

Another frequently quoted figure, 36,766, is also legitimate—but it measures robots ordered throughout North America in 2025, not robots installed in the United States. Mixing orders, installations and operational stock is how a real data point turns into statistical soup. (Association for Advancing Automation, 2026)

What the 3.3-jobs-per-robot study actually found

Economists Daron Acemoglu and Pascual Restrepo examined U.S. exposure to industrial robots from 1990 through 2007. Between 1993 and 2007, the operational stock increased by approximately 120,000 robots. Their nationwide model estimated an employment reduction of about 400,000 jobs—roughly 3.3 workers per additional robot. In the directly exposed local labor market, the estimated effect was closer to six workers per robot. Nationally, one additional robot per 1,000 workers was associated with a 0.2-percentage-point reduction in the employment-to-population ratio and a 0.42 percent wage reduction. (Acemoglu and Restrepo, 2020)

That 3.3 figure is an econometric result from a particular historical period, not a manufacturer's conversion chart. A robot installed to replace a retired robot is not another net robot added to the economy. A robot also can suppress hiring or shift a worker into another position without producing a same-day layoff.

Period	Robot measure	Employment calculation
1993-2007	About 120,000 added to operational stock	About 400,000 jobs; study estimate
2008-2019	About 293,388 installations	About 968,180 at 3.3; scenario only
2020-2025	About 215,137 installations	About 709,952 at 3.3; scenario only
2008-2025	About 508,525 installations	About 1,678,133 at 3.3; scenario only

Calculation note: Historical annual IFR installations are summed across 2008-2025. Because the series includes replacements and preliminary figures, the 3.3 multiplication is a transparent scenario—not a measured loss count. (IFR, 2017; IFR, 2018; IFR, 2021; IFR, 2022; IFR, 2023; IFR, 2025; IFR, 2026)

For the equipment-based ledger used in this article, the peer-reviewed 400,000 estimate for 1993-2007 is combined with the post-2007 mechanical scenario of 1,678,133. That produces 2,078,133 robot-related job-equivalents. It is the calculation requested and it is mathematically correct; its limitation is that later installations are not the same thing as net additions to operational stock.

Manufacturing lost millions—automation's share is disputed

The actual manufacturing-employment decline is not small. The BLS payroll series shows 17.797 million manufacturing jobs in January 1990, 17.284 million in January 2000, 11.447 million in January 2010, 12.745 million in January 2020 and 12.611 million in July 2026. From January 1990 through July 2026, manufacturing employment declined by 5.186 million, or 29.1 percent. (BLS/FRED, 2026)

Period	Starting employment	Ending employment	Change
1990-2000	17.797 million	17.284 million	-513,000
2000-2010	17.284 million	11.447 million	-5,837,000
2010-2020	11.447 million	12.745 million	+1,298,000
2020-Jul. 2026	12.745 million	12.611 million	-134,000

A Ball State University study calculated 5,647,700 manufacturing jobs lost during the early-2000s decline and attributed 87.8 percent of that loss to productivity growth. Applying its percentage gives approximately 4,958,681 positions attributed to productivity. That is the basis for the broader automation figure used later in this article. (Hicks and Devaraj, 2015)

But productivity is not synonymous with robots. It can include computerized machinery, process redesign, better logistics, software, outsourcing and changes in how output is measured. Susan Houseman of the Upjohn Institute argues that widely cited manufacturing-productivity numbers were distorted by the computer and electronics sector and therefore cannot explain the entire employment collapse. Trade, offshoring, recessions and changing domestic demand also matter. (Upjohn Institute, 2018)

The honest conclusion: The 4.96 million figure is a broad productivity/automation attribution from one influential study. It should be presented beside the competing interpretation—not advertised as a universally accepted robot-loss count.

Self-checkout and restaurant kiosks: the labor math

Self-service automation is easier to see than factory automation because customers perform the task themselves. The machine does not need to replace every cashier to reduce labor demand. Four cashier positions removed during every operating shift can erase more than four annual jobs because one continuously staffed position requires multiple employees across the year.

This analysis assumes two eight-hour shifts per day, 365 days per year and a 2,080-hour full-time-equivalent work year. One position staffed for both daily shifts therefore equals 2.8077 annual FTEs. It then applies the requested 50 percent replacement rule: two restaurant kiosks equal one position per shift, and eight retail self-checkouts equal four positions per shift.

Category and assumption	Gross annual FTE	Net annual FTE
Restaurant kiosks: about 75,000 two-kiosk site-equivalents	210,577	210,577
Home Depot + Lowe's: 3,794 locations, eight SCOs assumed	42,610	31,957
All supermarkets: 45,575 stores, eight SCOs assumed	511,842	383,882
Restaurants + home improvement	253,186	242,534
Broad model including every supermarket	765,029	626,416

Net definition: For an eight-machine self-checkout bank, one attendant per shift is deducted from the four-position gross reduction, leaving three net positions per shift. Restaurant kiosks receive no attendant deduction because their remaining workers generally perform food-production, fulfillment and customer-service duties rather than staffing each kiosk. Totals are calculated from unrounded formulas and then rounded; separately adding displayed rows can therefore differ by one FTE.

Datos Insights reported that the global installed base of restaurant self-ordering kiosks was approaching 450,000, with the United States home to close to one-third. That supports an inference of approximately 150,000 U.S. terminals, or 75,000 two-kiosk site-equivalents. It is not an official count of 75,000 restaurants with exactly two machines. (Datos Insights, 2026)

The Home Depot reported 2,035 U.S. stores at the end of fiscal 2025. Lowe's reported 1,759 U.S. stores and outlets, producing a combined 3,794 locations. Neither filing says that every store has exactly eight self-checkouts; eight is the stated model assumption based on observed configurations. (Home Depot, 2026; Lowe's, 2026)

FMI reports 45,575 U.S. supermarkets in 2024 and says self-checkout handled 33 percent of supermarket transactions in 2025. There is no public national terminal census proving eight machines at every supermarket. Therefore, the grocery calculation is the broad scenario, not a verified store-by-store total. (FMI, 2026)

The attendant adjustment is supported by actual staffing practices. In an ECR Retail Loss survey, the most common ratio was one employee supervising five or six machines; 38 percent of respondents supervised seven or more. BLS projects cashier employment to decline by 313,600 jobs, or 10 percent, between 2024 and 2034 because of self-service checkout and online sales. That BLS projection overlaps this model and must not be added to it. (ECR Retail Loss, 2022; BLS, 2025a)

ATMs and electronic tolling show two different outcomes

The ATM Industry Association estimated 520,000 to 540,000 active ATMs in the United States at the end of 2022 and acknowledged that an exact count is virtually impossible. There is no defensible national jobs-per-ATM coefficient. (ATMIA, 2023)

ATMs reduced the average number of tellers needed in an urban branch from approximately 20 to 13. But lower branch-operating costs encouraged banks to open more locations; urban branches increased 43 percent during the historical expansion. The machines reduced labor per branch without initially eliminating teller employment nationwide. The later picture is different: BLS projects teller employment to fall from 347,400 in 2024 to 302,500 in 2034, a reduction of 44,900 jobs or 13 percent, citing online banking, video kiosks and enhanced ATMs. (Bessen, 2015; BLS, 2025b)

Electronic tolling provides a rare direct layoff count. When the Pennsylvania Turnpike made all-electronic toll collection permanent in 2020, it laid off 492 Fare Collection and Ticket Systems Audit employees. COVID-19 accelerated the timetable, but the permanent automated system eliminated the need for those cash-collection positions. (Pennsylvania Turnpike Commission, 2023)

Ten percent autonomous trucking: what it would mean

BLS counted 2,235,100 heavy and tractor-trailer truck-driver jobs in 2024 and projects 2,324,400 in 2034, an increase of 89,300 before any assumed autonomous shock. Median annual pay was $57,440. (BLS, 2025c)

Ten percent of the 2024 workforce is 223,510 positions. Ten percent of the official 2034 projection is 232,440. Extending BLS's four-percent decade growth rate two additional years produces an illustrative 2036 base of approximately 2.343 million positions; ten percent equals 234,281. Spread evenly across ten years, that is approximately 23,428 positions or foregone hires per year, representing about $13.46 billion in annual wages at the 2024 median.

But ten percent of trucks equipped for autonomous operation is not automatically ten percent of driver jobs eliminated. A hub-to-hub model retains human drivers for local pickup, delivery, inspections and first- and last-mile work. The 234,281 figure applies only if ten percent of total heavy-truck driving work becomes genuinely driverless.

GAO reviewed speculative estimates ranging from fewer than 300,000 to more than 900,000 driver jobs affected over ten to twenty years and emphasized the uncertainty. Meanwhile, FHWA expects combination-truck mileage to grow about one percent annually through 2043, so increased freight demand could offset part of the labor reduction. (GAO, 2019; FHWA, 2025)

What total can reasonably be presented?

A single number can be useful only when its ingredients are shown. The following totals use alternative accounting methods; they are not meant to be added together.

Accounting approach	Included components	Result
Equipment-based core	Robot ledger; restaurant and home-improvement net FTE; PA toll layoffs; 10% trucking	2,555,440
Equipment-based broad net	Core plus all-supermarket net scenario	2,939,321
Equipment-based broad gross	Robot ledger; all self-service gross FTE; PA toll layoffs; 10% trucking	3,077,934
Broad productivity lens	Ball State manufacturing attribution; broad self-service net; PA toll layoffs; 10% trucking	5,819,870

Rounding note: Each grand total is calculated from unrounded inputs and rounded only at the end, so subtracting displayed totals can differ by one from a separately rounded component.

The approximately 2.94 million figure is the corrected total for the equipment-based model developed here. The approximately 5.82 million figure substitutes the broader Ball State productivity attribution for the robot-specific ledger; adding both would double-count manufacturing effects.

A separate estimate of 51 million to 102 million work-equivalent roles comes from applying a 30-to-60-percent automation range to a roughly 170-million-person workforce. That answers a different question—how much work automation may currently perform or affect—not how many identifiable people have lost jobs since the 1990s. Presenting it as historical layoffs would be wrong.

The real lesson is hidden in the method

Automation does not have to eliminate an occupation overnight to remove jobs. It can reduce the number of workers required per shift, eliminate replacement hiring, transfer work to customers, suppress wage growth, move people into lower-paying work and concentrate the gains from higher productivity among owners and technology providers. The absence of one giant national layoff announcement is not evidence that the labor effect is imaginary.

At the same time, inflated claims are not helpful. Robot installations cannot be treated automatically as net additions. Every productivity improvement cannot be assigned entirely to automation. Ten percent autonomous capability cannot be assumed to equal ten percent driver elimination. The strongest argument is the one that labels each figure honestly and shows the arithmetic.

On that basis, the defensible public statement is straightforward: the corrected equipment-based model produces approximately 2.94 million net job/FTE-equivalents, while a broader and disputed productivity-based interpretation reaches approximately 5.82 million. The country may debate whether those losses are an acceptable price for productivity. It should stop pretending the losses do not exist.

Annual U.S. industrial-robot installation ledger

This appendix shows every annual installation figure used in the post-2007 robot scenario. The right column is simple arithmetic—annual installations multiplied by 3.3 and rounded to the nearest whole job-equivalent. It is not an official annual layoff count.

Year	U.S. installations	Installations × 3.3
2008	13,380	44,154
2009	6,831	22,542
2010	14,380	47,454
2011	20,555	67,832
2012	22,414	73,966
2013	23,679	78,141
2014	26,202	86,467
2015	27,504	90,763
2016	31,500 preliminary	103,950
2017	33,192	109,534
2018	40,373	133,231
2019	33,378	110,147
2020	30,787	101,597
2021	34,987	115,457
2022	39,576	130,601
2023	37,587	124,037
2024	34,200	112,860
2025	38,000 preliminary	125,400
Total	508,525	1,678,133

Source note: IFR public releases for 2008-2025. The 2008-2016 chart and the 2025 release explicitly identify preliminary data; later IFR publications revise some past observations. The total above reproduces the published series used in this article and should be read as an auditable scenario input, not a count of net additions to robot stock.

Sources

Acemoglu, Daron, and Pascual Restrepo. "Robots and Jobs: Evidence from U.S. Labor Markets." Journal of Political Economy, 2020. View source

Association for Advancing Automation. "A3 Robotics Statistics: 2025 North American Orders." A3, 2026. View source

ATM Industry Association. "ATM Population in the U.S. Shows Strong Growth Over the Past Year." ATMIA, September 6, 2023. View source

Bessen, James. "Toil and Technology." Finance & Development, International Monetary Fund, March 2015. View source

Bureau of Labor Statistics. "Cashiers: Occupational Outlook Handbook." U.S. Department of Labor, 2025. View source

Bureau of Labor Statistics. "Tellers: Occupational Outlook Handbook." U.S. Department of Labor, 2025. View source

Bureau of Labor Statistics. "Heavy and Tractor-trailer Truck Drivers: Occupational Outlook Handbook." U.S. Department of Labor, 2025. View source

Bureau of Labor Statistics / Federal Reserve Bank of St. Louis. "All Employees, Manufacturing (MANEMP)." FRED, updated August 7, 2026. View source

Datos Insights. "Hunger for Kiosks Grows as the Technology Extends Beyond Fast Food." Datos Insights, February 19, 2026. View source

ECR Retail Loss. "Self-checkout Management: Working on the 'Front Line'." ECR Retail Loss, November 30, 2022. View source

Federal Highway Administration. "2025 FHWA Forecasts of Vehicle Miles Traveled." U.S. Department of Transportation, October 10, 2025. View source

FMI - The Food Industry Association. "Food Industry Facts." FMI, accessed August 13, 2026. View source

Government Accountability Office. "Automated Trucking: Federal Agencies Should Take Additional Steps to Prepare for Potential Workforce Effects." GAO-19-161, March 7, 2019. View source

Hicks, Michael J., and Srikant Devaraj. "The Myth and the Reality of Manufacturing in America." Ball State University, hosted by NIST, 2015. View source

The Home Depot, Inc.. "Fiscal 2025 Form 10-K." U.S. Securities and Exchange Commission, 2026. View source

International Federation of Robotics. "Global Industrial Robot Sales Doubled Over the Past Five Years." IFR, October 18, 2018. View source

International Federation of Robotics. "Executive Summary: World Robotics 2021 Industrial Robots." IFR, 2021. View source

International Federation of Robotics. "All-Time High with Half a Million Robots Installed in One Year." IFR, October 13, 2022. View source

International Federation of Robotics. "World Robotics 2023 Report: Asia Ahead of Europe and the Americas." IFR, September 26, 2023. View source

International Federation of Robotics. "U.S. Lags China in Factory Robot Deployment by 5 to 1 Ratio." IFR, September 25, 2025. View source

International Federation of Robotics. "U.S. Robot Industry Returns to Double Digit Growth." IFR, June 18, 2026. View source

International Federation of Robotics. "U.S. Industrial Robot Installations, 2008-2016." IFR Statistical Department, 2017. View source

Lowe's Companies, Inc.. "Fiscal 2025 Form 10-K." U.S. Securities and Exchange Commission, 2026. View source

Pennsylvania Turnpike Commission. "All-Electronic Tolling Layoff." Official Statement, 2023. View source

Upjohn Institute for Employment Research. "Understanding the Decline in Manufacturing Employment." W.E. Upjohn Institute, 2018. View source