Skip to sources
Time Machine

Data Entry Keyers

Scrub through 146years of this role's history, from when it first emerged, through every wave of technology that reshaped it, to the cited projections for where it's heading next.

2026drag to travel through time
19001925195019752000now
Country
2026
Known today as Data Entry Keyers (BLS SOC 43-9021)
Latest actual · 2024
135K
OEWS is a point-in-time survey snapshot, not a continuous time series; BLS advises against using it for year-over-year trend comparison.
Latest actual · 2024
$39,850
Source: BLS-OEWS
Each dot is a cited figure over time; the dotted line only links them (values between aren't measured). Hollow dots are estimates.
Beat · 2025

The WEF Future of Jobs Report 2025 lists data entry clerks as the third-fastest-shrinking occupational category globally through 2030, projecting a 34% decline behind only postal service clerks (-40%) and bank tellers (-35%). The WEF survey of over 1,000 employers identifies AI and automation as the primary drivers. Separately, BLS projects the U.S. occupation specifically to decline 25.9% from 141,600 in 2024 to 104,900 by 2034. The gap between the WEF figure (faster) and BLS (slower) reflects the difference between employer expectations about AI adoption speed and BLS's more conservative productivity-model methodology.

Tools of the era

The tools that defined the work

Select an era to see how it reshaped the work.

  • Hollerith pantograph punch + tabulating machine (census and insurance era)

    Hollerith's pantograph punch required an operator to align a census schedule beside the punch, read each item, and press the appropriate hole position using a hand-held stylus guided through a template. The punch produced no printed output on the card: the operator had to verify work by re-reading the card positions visually. Tabulating machines downstream read the holes electrically and incremented mechanical counters. The entire system was designed for batch statistical counting, not for individual record lookup, and the operator's job was purely transcription: copy the paper into the machine-readable medium as accurately as possible.

    Effect on the work

    For the 1890 census, Hollerith's machines and their operators completed the population count in six months, compared to seven years for the 1880 census done by hand. The Census Bureau estimated the system saved over $5 million versus the manual alternative. This demonstrated conclusively that a modest number of trained operators with machines could outperform a large manual clerical force.

    Punch-card systemsBatch accounting
  • IBM 026 Printing Card Punch (and Remington Rand / UNIVAC keypunch era)

    The IBM 026 Printing Card Punch, introduced in 1949 and widely deployed through the 1950s and 1960s, became the defining tool of the keypunch operator profession. Unlike earlier punches, the 026 printed the character it punched along the top edge of the card, giving operators immediate visual feedback and making verification easier. A program drum could be loaded with a customized column-skip-and-duplicate routine for specific form layouts, allowing an expert operator to process 200 cards per hour. Employers set up keypunch departments of dozens or hundreds of operators working in parallel, a factory-style arrangement that made data entry a mass-production clerical activity. Remington Rand's competing UNIVAC keypunches, designed for their own systems, added magnetic tape output capability as early as the mid-1950s via the UNITYPER, enabling direct-to-tape entry that bypassed the card entirely in high-volume environments.

    Effect on the work

    By the mid-1960s, keypunch operator was among the fastest-growing occupational categories in BLS clerical surveys. The occupation grew from a handful of Census Bureau workers in 1890 to an estimated 75,000 in 1950 and to roughly 291,000 by 1980, driven entirely by the spread of batch data-processing into banking, insurance, retail, government, and health care.

    Work toolChanging equipment
  • Video display terminals (VDTs) and online data entry (replacing punched cards)

    The punched card had a fundamental workflow problem: operators created cards in a batch, the cards were carried to a card reader, the reader submitted them to a mainframe, and errors were discovered only after the job ran. Video display terminals changed this. An operator typing at a VDT could see the data on screen, receive immediate validation feedback from the application, and correct errors in real time. IBM's 3270 terminal family (1971) and the proliferation of minicomputer-based systems through the 1970s drove a rapid shift away from card-based batch entry. By the early 1980s, most large data-processing installations had moved to online VDT entry; the keypunch machine was becoming a legacy device. The occupational title was slowly changing from "keypunch operator" to "data entry operator," reflecting the shift from a specific machine to a generic task.

    Effect on the work

    The VDT transition did not reduce employment in the short term: if anything, the faster throughput and broader application of online systems expanded demand for data entry operators through the 1970s and into the early 1980s. Employment peaked in the first half of the 1980s and then began a structural decline as personal computers let originating departments enter their own data directly, bypassing the dedicated data-entry pool.

    Work toolChanging equipment
  • Personal computers and networked business applications (PC-era data entry)

    The IBM PC (1981) and its successors did not immediately eliminate data entry jobs, but they restructured who did the work and why a dedicated specialist was still needed. By the late 1980s, professional and managerial workers were entering their own correspondence and spreadsheets on PCs. What remained for dedicated data entry keyers was the volume work: batch-processing incoming forms, scanning paper records into database systems, and transferring data between incompatible systems where no automated bridge existed. Networked PC applications through the 1990s created enormous new demand for this kind of bridging work: health care, financial services, and government agencies accumulated mountains of paper records that needed to migrate into digital systems, and the data entry workforce absorbed much of that backlog. The 1990s thus saw a second wave of employment growth, as the digitization backlog from decades of paper-based record-keeping created new demand even as the front-end data creation shifted to originating departments.

    Effect on the work

    Employment of data entry keyers grew through the 1990s into the early 2000s, reaching roughly 330,000 by 2004, as the digitization of healthcare records, legal documents, financial forms, and government records created a sustained demand for manual transcription of legacy paper into digital systems. This was the "digitization backlog" wave: not new data creation, but migration of old paper into new systems.

    Work toolChanging equipment
  • OCR and forms-processing software (first wave of systematic automation)

    Optical character recognition technology had existed since the 1950s but only became accurate and affordable enough for large-scale commercial deployment in the late 1990s and 2000s. Software from vendors like Kofax (founded 1985, enterprise OCR deployments from the mid-1990s), ReadSoft, and ABBYY FineReader automated a significant share of structured-form processing: printed invoices, standardized government forms, and typed medical records could now be captured with 90-95% accuracy by machine. Human data entry keyers were increasingly repositioned as quality-control reviewers who corrected the minority of fields that OCR got wrong, rather than entering the full document from scratch. Barcode scanning at retail and logistics eliminated data entry from those sectors. The combination of these factors drove employment from 330,000 in 2004 to approximately 234,700 by 2010, a 29% drop in six years.

    Effect on the work

    Employment fell from approximately 330,000 in 2004 to 234,700 in 2010. OCR, barcode scanning, and forms-processing software eliminated roughly a third of the job population in six years. The remaining workforce concentrated in sectors with high proportions of handwritten, non-standard, or regulated documents where OCR accuracy remained insufficient.

    Work toolChanging equipment
  • Intelligent document processing and AI extraction (Rossum, Hyperscience, UiPath)

    Intelligent document processing (IDP) platforms, which combine deep-learning OCR with field-classification models and structured extraction pipelines, represent a qualitative leap beyond first-generation OCR. Where earlier OCR read printed text character by character, IDP platforms from vendors including Rossum (2017), Hyperscience (founded 2014), Nanonets (2017), and UiPath Document Understanding (2019) learn to locate and extract specific fields from documents they have never seen before, handle semi-structured layouts, and route low-confidence extractions to a human review queue rather than silently accepting errors. The practical effect is that IDP can automate 80% or more of the fields on a typical invoice or form batch while handing the exception queue to a human reviewer. The data entry keyer's role in this model is not pure transcription but adjudication: reviewing, correcting, and approving the fields that the machine flagged as uncertain. Generative AI tools further extended the automation frontier in 2023-2024 by enabling free-form extraction from unstructured documents and multi-step workflow automation that connects extraction directly to downstream ERP and CRM systems.

    Effect on the work

    Employment fell from approximately 375,000 in 2014 to 141,600 by 2024, a decline of roughly 62% in a decade. The IDP and AI extraction wave is projected to eliminate a further 25.9% of remaining positions by 2034, according to BLS projections. The WEF Future of Jobs Report 2025 lists data entry clerks as among the fastest-declining occupations globally, projecting a 34% decline by 2030.

    Work toolChanging equipment
Projection cone · present → 2034

What credible sources project

Scrub the slider past now to anchor each scenario on the scrubber. The spread is the range of futures credible sources project for this role.

Employment outlook
Projected change in the number of people doing this work.
BLS National Employment Matrix 2024-34
2034
-25.9%
BLS employment projections model: industry-occupation matrix driven by output and productivity assumptions. The 2024-34 cycle projects data entry keyers (43-9021) to decline from 141,600 in 2024 to 104,900 by 2034, a loss of 36,700 positions (-25.9%). This is among the steeper percentage declines in the entire BLS occupational matrix. The projection assumes continued adoption of IDP platforms, AI-assisted extraction, and ongoing outsourcing of back-office data work. It does not model the pace of AI generative-extraction adoption, which could accelerate the decline beyond BLS's baseline.
WEF Future of Jobs Report 2025
2030
-34%
WEF employer survey of 1,000+ companies across 22 industry clusters and 55 economies, asking respondents to estimate occupation-level headcount changes by 2030 driven by technology, demographic, and economic factors. The WEF 2025 report lists data entry clerks as among the five fastest-shrinking job categories globally, with a projected -34% decline by 2030. Only postal service clerks (-40%) and bank tellers (-35%) score more steeply. The WEF projection is faster than the BLS 2024-34 figure because it captures global enterprise expectations about AI adoption speed rather than BLS's more conservative productivity-modeling approach.
AI task exposure
Share of the role’s tasks that researchers estimate AI can do. This is a measure of task exposure, not a forecast of jobs lost.
Eloundou et al. "GPTs are GPTs" (2023)
2028
81%
of tasks
GPT-4 task-by-task LLM-exposure labeling on O*NET tasks. Data entry keyers score in the highest exposure tier: their core tasks (entering data from source documents, verifying entered data, performing clerical functions) are exactly the structured transcription tasks that large language models handle with high accuracy. Eloundou et al. found that the tasks of data entry workers are among the most exposed to LLM-assisted automation in the entire O*NET task database. The 81% figure represents the share of task-weighted work hours estimated to be directly exposable to LLM augmentation or automation, not a projected headcount loss; actual employment change depends on adoption speed and organizational restructuring, which are captured in the BLS and WEF figures above.
Today, in this role

What's shifting in the work right now

The historical view above shows how this role has moved. This is the present-day detail: which AI tools are picking up which tasks, where the edge still is, and the natural directions this work can grow.

What's changing in your day

Three parts of your work where AI is already doing real lifting, and what stays yours.

AI is sitting alongside you hereReview batches of invoices, purchase orders, or medical records in an IDP platform (such as Rossum or Nanonets) and adjudicate low-confidence AI extractions before routing to downstream systems.

Review batches of invoices, purchase orders, or medical records in an IDP platform (such as Rossum or Nanonets) and adjudicate low-confidence AI extractions before routing to downstream systems.[6],[7]

Where your edge is

Build proficiency in at least one IDP review interface; learn to tune extraction confidence thresholds and flag systematic misreads to the vendor support queue.

AI is sitting alongside you hereVerify extracted records against source documents using cross-field validation rules (e.g., totals reconciliation, date-range checks, required-field completeness) before data is committed to an ERP or CRM system.

Verify extracted records against source documents using cross-field validation rules (e.g., totals reconciliation, date-range checks, required-field completeness) before data is committed to an ERP or CRM system.[8],[4]

Where your edge is

Learn SQL or spreadsheet pivot-table techniques to run exception reports that surface systematic extraction errors across a batch, shifting from per-record checking to population-level QA.

AI is sitting alongside you hereManually key data from handwritten forms, legacy paper documents, or non-standard source materials that OCR and AI extraction tools cannot process reliably.

Manually key data from handwritten forms, legacy paper documents, or non-standard source materials that OCR and AI extraction tools cannot process reliably.[1],[4]

Where your edge is

Specialize in high-stakes document categories (clinical handwritten notes, court filings, non-Latin scripts) where human transcription remains necessary and carries legal accountability.

Where this role is heading

Natural next steps for someone with your foundation: not exits, evolutions.

A direction you could grow

Computer User Support Specialists

Data entry keyers who participate in IDP or RPA automation projects accumulate working knowledge of software configuration, exception triage, and user training that maps well onto a support specialist role. The pivot requires broadening technical troubleshooting skills but the software fluency is already present.

What you'd add
  • · CompTIA A+ or equivalent hardware/OS fundamentals
  • · Help-desk ticketing systems (Jira Service Management, Freshdesk)
  • · RPA tool basics (UiPath Community Edition free training)
  • · IT service management (ITIL foundations)
What it takesSome new skills to pick up
Share this year
Drops anyone you send it to straight into 2026.
Preview card
Part of Admin & Office · see all 39roles →
Different role?

See the same long-arc view for your own profession.

Browse the directory by industry, or search by title or SOC code. New roles ship every few weeks. Every profile cites every claim.

Browse all roles

The data behind this timeline

On record since1890
Latest tracked employment135,280 (US, 2024)
Latest median pay$39,850 (2024)
Outlook-34% by 2030 (WEF Future of Jobs Report 2025)
View all 26 cited data points
YearUS employmentMedian annual paySource
1890200n/aESTIMATE
195075,000n/aESTIMATE
1960n/a$3,800ESTIMATE
1980291,000n/aESTIMATE
2003339,010$22,600BLS-OEWS
2004313,590$23,250BLS-OEWS
2005296,700$23,810BLS-OEWS
2006295,650$24,690BLS-OEWS
2007286,540$25,370BLS-OEWS
2008272,810$26,120BLS-OEWS
2009243,550$27,150BLS-OEWS
2010219,530$27,450BLS-OEWS
2011211,200$27,690BLS-OEWS
2012207,280$28,010BLS-OEWS
2013207,660$28,470BLS-OEWS
2014205,950$28,870BLS-OEWS
2015199,240$29,460BLS-OEWS
2016194,810$30,100BLS-OEWS
2017180,100$30,930BLS-OEWS
2018174,930$32,170BLS-OEWS
2019159,930$33,490BLS-OEWS
2020151,520$34,440BLS-OEWS
2021147,170$35,630BLS-OEWS
2022157,380$36,190BLS-OEWS
2023154,230$37,790BLS-OEWS
2024135,280$39,850BLS-OEWS
Embed this timeline on your site

Free for any site. Paste this where the timeline should appear; it stays interactive, every datapoint stays cited, and it sets no cookies on your page. How embedding works

<iframe src="https://futurehistory.earth/embed/43-9021"
  width="100%" height="430" style="border:0"
  title="Data Entry Keyers, a Future History timeline"
  loading="lazy"></iframe>

See all roles in Admin & Office