LO
Get told when a verified record lands on this occupation → · Mark which of these tasks are yours (VOLO Pro, free during the launch) →
Data entry clerk
Keys data from paper forms, scans and documents into systems, checks it against the source, and fixes what does not match. US projections expect this occupation to shrink by about a quarter by 2035, and machines now read much of what used to be keyed — census forms, tax returns, invoices, financial statements. But the machines route what they cannot read with confidence back to people, a large IRS digitisation effort has so far processed only a small share of paper returns, and firms that automate data capture keep people in the loop for exceptions and quality.
Ask your supervisor what share of documents your team now receives already extracted by software, and who checks them.
This is not a probability of losing your job. It combines how much of the role's task load is exposed to automation with how far adoption has actually gone — useful for comparing occupations on one consistent basis, and for nothing else.
Written for people whose main job is keying and checking data — data entry clerks and keyers in back offices, public agencies and outsourcing firms. Administrative assistants, receptionists and medical coders have their own pages. The evidence is a US employment projection, an inspector general's audit of the IRS, a Census Bureau evaluation, annual reports from a data company and an outsourcing firm, and two extraction benchmarks; it establishes where machines read documents and where people remain, not why the projected decline happens.
What is actually changing#
The unit of analysis is the task, not the job title. A role is not replaced — its task mix shifts.
Each tile is one task. Its size is how much of the job it is; its colour is where the task is heading. Click a tile to see what the judgement does not establish.
A company's description of its own operations, one agency's audit and an older test; none counts how many keying jobs have gone.
Benchmarks, one by a company that sells document products; they do not measure how much checking work remains in practice.
A firm describing the service it sells and an older government test; how many exception-handling roles exist is not measured.
A vendor's claim about its own service and one audit; neither counts mailroom jobs.
Is this your job? Say so and this page narrows to your share of it.
A job title is a bundle of tasks bought together, and no two people hold the same bundle. Nothing is sent anywhere — it stays in this browser.
Read all 4 tasks in full — direction, reasoning and limits →
Recent changes#
United States. The projections table lists data entry keyers (43-9021) at 131,800 jobs in 2025 and 98,200 in 2035, a change of −33,600 or −25.5 percent, with about 7,700 openings a year. The table gives no reason for the decline; the projection counts jobs in one country and does not attribute the change to AI.
A named person with standing publicly predicted something, on a date, in an attributable statement. It is recorded so that who said what, and when, stays checkable — and it never moves a task's assessment, because a prediction is not an observation. Its value arrives later: the record sits on the same page as the evidence about that occupation, so anyone reading the forecast reads the record of what happened next beside it. That is the reckoning; this site publishes no verdict on whether a forecast came true.
A business-process outsourcing company's annual report. It says its intelligent document processing platform, paired with dedicated human-in-the-loop exception handling, converts correspondence into electronic data and reduces the need for dedicated mailroom personnel who open, review and sort documents, and that in highly regulated industries automated workflows still require dedicated human oversight, with its personnel handling complex exceptions, sensitive data and quality control. It is the company describing the service it sells.
A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
A financial data company's annual report. It says it continued to deploy and enhance AI-powered data collection and ingestion processes with independent standards and human-in-the-loop governance focused on coverage, reliability and processing speed while maintaining data quality. It is the company describing its own operations; it does not link any change in headcount to AI.
An employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.
A benchmark of extracting structured data from 35 PDF documents with human-annotated gold labels, 12,867 fields in all. The authors report that frontier models remain unreliable on realistic schemas, that performance degrades sharply with schema breadth, culminating in 0% valid output on a 369-field financial reporting schema across all tested models, and that on one domain models achieve 90% valid output but only a 12.5% pass rate. The authors work at a company that sells document products; it is a benchmark, not a measure of practice.
A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
United States, the IRS. The audit says pilots in which contractors scanned paper returns and extracted the data proved that paper returns can be digitised, but only about 3.8 million (7 percent) of 53.3 million paper-filed Forms 940, 941 and 1040 received from February 2023 through December 2024 were processed this way; as of May 2025 contractors had scanned nearly 517,000 (5 percent) of 9.8 million received in the 2025 filing season. In September 2025 the IRS awarded digitisation contracts totalling $2.3 billion through fiscal year 2030, and contractors must recruit staff to receive and open mail and review return data for accuracy. Work moving to contractors is not the same as work being automated.
Small-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.
A benchmark of extracting fields from invoices and receipts with language models, zero-shot, on open datasets. The authors report that Gemini 2.5 Pro achieved the highest accuracy across all three datasets: 87.46% on scanned receipts, 96.50% on clean invoices and 92.71% on scanned invoices. It measures field accuracy on public datasets, not use in accounts-payable departments.
A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
United States, the Census Bureau's processing centre, where clerks key all write-in responses from scanned questionnaires. In a test across more than 71,000 values, character recognition matched the keyed value about 73 percent of the time; almost 27 percent it did not read, and when it cannot determine a field's value the field goes to a keyer. Of all values it read, 0.3 percent were in error, less than the 1 percent error rate required for keyed batches. It is an older test on numeric fields only.
Small-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.
What this means for you#
If you are considering this work, expect fewer pure keying jobs, and aim for the roles that remain around the machine: checking records, handling exceptions and sensitive data, and knowing the rules the data must follow.
Expect the volume you key by hand to fall as documents are read by software, and your work to move towards reviewing what it gets wrong. Knowing the domain behind the data — tax, claims, finance — is what makes you the person who handles exceptions.
Your options#
Four directions, each with its real constraints and one thing you can test this week. Continuing as you are is a legitimate choice — it just has to be a chosen one.
Move from keying to checking and exceptions
Automated capture sends what it cannot read, and what must be checked, to people.
There are fewer of these roles than keying roles, and they expect domain knowledge.
Ask your supervisor what share of documents your team now receives already extracted by software, and who checks them.
Move into a domain role that uses the data
Knowing what the data means — in claims, billing or accounts — carries into roles that decide on it.
These roles often ask for training or certification.
Pick one document type you key most and list the decisions someone downstream makes with it.
Stay where paper and sensitive data remain
Public agencies still receive large volumes of paper, and digitisation has been slow.
The work is likely to move to contractors and shrink as digitisation catches up.
Find out whether your employer has a digitisation contract or plan, and when it is due.
Common questions#
Much of the keying, yes, over time; the checking, not yet. US projections expect data entry keyer employment to fall 25.5 percent from 2025 to 2035, and machines now read census forms, tax returns and financial documents. But extraction benchmarks show models still make errors that need review, and firms that automate capture keep people for exceptions and quality.
We do not answer that with a number of years. Watch how much of your team's work arrives already extracted by software, and whether your employer has signed a digitisation contract. Those tell you more about your own job than any date.
The US projection does not say. It expects data entry keyer employment to fall from 131,800 to 98,200 between 2025 and 2035 but gives no cause, so we do not attribute the decline to AI.
The ones machines hand back: checking records against sources, handling documents software cannot read, and knowing the rules of the domain the data comes from.
What these judgements rest on#
2 of 4 task judgements on this page are backed by a verified event and 2 are platform inference, each labelled where it appears. Behind them sit 3 technology dimensions, a reconstructed trajectory since language models reached the public, and 7 verified events.
See which technologies, how it got here, and the method →
Where it sits in the official classification: skills, knowledge, related jobs →
Other roles in the same function#
A company divides its work into functions before it divides it into jobs. These sit in Administration alongside this one — a fact about org charts, not a judgement that they are similar or that they are changing in the same direction.
Administrative assistant · Government service clerk · Receptionist / front desk · Medical coder