UK AI Exposure · Administrative and secretarial occupations
Data entry administrators
Data entry administrators enter a variety of information into databases using various software packages and assist colleagues in retrieving information.
- Employees (UK)
- 16k
- Median annual pay
- £26,534
- Exposure score ?
- 7.9/10 High 9.1/10 Very high strict reading · with tools is 9.1/10 with-tools reading · strict is 7.9/10
- Wage exposure
- £335m £386m
Higher exposure than 100% of the 379 UK occupations we scored.
What this score means
Most of the routine task inventory in this role can already be done by a capable LLM. That doesn't mean the role disappears - it means the shape changes, and one person can credibly do the work of several.
If you're in this role, here's what to do now
Stop doing anything an LLM can do. Your edge is judgment, relationships, taste, and the parts of the work that require you to be in the room. The operators who notice this first and redesign their workflow around it will be paid for those things; the ones who cling to the old task list will compete against AI at AI's prices.
Almost every routine task in this role is within reach of today's language models. Roles at this level are getting rebuilt - often not by disappearing, but by one person using AI to do three or five people's output.
If you're in this role, here's what to do now
You don't need to be afraid. You need to be the person doing the rebuilding. The operators who learn to direct AI at scale in this kind of work become hugely valuable. The ones who wait to be told what to do get told what to do - and that thing is often 'we don't need as many of you anymore.'
Where a project with Alex usually starts for this role
These are the highest-importance tasks a language model can already handle directly today. In a typical engagement the first wins come from building workflows around these, so they stop eating your team's time.
-
Locate and correct data entry errors, or report them to supervisors.
-
Compile, sort, and verify the accuracy of data before it is entered.
-
Compare data with source documents, or re-enter data in verification format to detect errors.
These are the highest-importance tasks AI can already handle when paired with the right tools and context. In a typical engagement the first wins come from building workflows around these — usually the difference between an LLM that can technically do the job and one that actually does it inside your business.
-
Locate and correct data entry errors, or report them to supervisors.
-
Compile, sort, and verify the accuracy of data before it is entered.
-
Compare data with source documents, or re-enter data in verification format to detect errors.
Every role has three or four wedges like these. Finding them takes an hour. Turning them into a workflow your team actually uses takes a few days. Talk to Alex about a project →
The full task breakdown
Every O*NET task for this occupation, split by what AI can already do unaided versus what still needs a human. Importance is O*NET's 1–5 rating of how central each task is to the role.
Tasks via O*NET "Data Entry Keyers" (43-9021.00).
What AI can already do
7 of 9 tasks · unaided
Locate and correct data entry errors, or report them to supervisors.
Compile, sort, and verify the accuracy of data before it is entered.
Compare data with source documents, or re-enter data in verification format to detect errors.
Store completed documents in appropriate locations.
Select materials needed to complete work assignments.
Maintain logs of activities and completed work.
Resolve garbled or indecipherable messages, using cryptographic procedures and equipment.
Where humans still hold the line
2 of 9 tasks
Read source documents such as canceled checks, sales reports, or bills, and enter data in specific data fields or onto tapes or disks for subsequent entry, using keyboards or scanners.
Load machines with required input or output media, such as paper, cards, disks, tape, or Braille media.
Tasks via O*NET "Data Entry Keyers" (43-9021.00).
What AI can already do
8 of 9 tasks · with tools
Locate and correct data entry errors, or report them to supervisors.
Compile, sort, and verify the accuracy of data before it is entered.
Compare data with source documents, or re-enter data in verification format to detect errors.
Read source documents such as canceled checks, sales reports, or bills, and enter data in specific data fields or onto tapes or disks for subsequent entry, using keyboards or scanners.
Store completed documents in appropriate locations.
Select materials needed to complete work assignments.
Maintain logs of activities and completed work.
Resolve garbled or indecipherable messages, using cryptographic procedures and equipment.
Where humans still hold the line
1 of 9 tasks
Load machines with required input or output media, such as paper, cards, disks, tape, or Braille media.
Stay on top of this
One email a week, written for people who aren't AI nerds. What's actually real, what's hype, and what smart operators are doing about it.
Get the weekly note
One email a week from Alex on how AI is changing UK work, how to get ahead of it, and what smart operators are actually doing. Written for people who aren't AI nerds.
Free. Unsubscribe any time.
Or go deeper:
Methodology
This role's exposure score comes from Eloundou et al's 2023 GPT task labels, aggregated by O*NET importance within each O*NET-SOC code, then bridged to UK SOC 2020 via ISCO-08 (ONS Vol 2 coding index) and US SOC 2010 (BLS crosswalk). Employment and median pay come from ONS ASHE Table 14.7a, 2025 provisional. ASHE covers employees only, so self-employed workers are not counted.
Get the weekly note. One email on how AI is changing UK work.