Will AI replace National government administrative occupations?

Not as a job, but it is already doing parts of the work. Across the 55 official task statements scored for National government administrative occupations (United Kingdom, SOC 4111), 51% of the importance-weighted core work is made of tasks today's AI could already do most of. The overall exposure score is 57 out of 100 (range 52–63, band: partial). That is a statement about tasks, not about headcount: this measures what AI could do, not whether any employer adopts it, whether the law allows it, or whether doing the routine parts faster creates more demand for the human parts. Figures are from release 2026-q4.1.

What should someone working in “National government administrative occupations” do about AI?

Start from the ledger rather than the headline: 51% of this job's weighted core work is exposed, and roughly 27% is not. The practical move is to spend more of your week on the tasks that score low, the ones above, and to get fluent at directing AI through the tasks that score high, because those are the parts that change whether or not you are ready for them. This page does not predict your job, and nothing here is career advice tailored to you: the score describes the occupation, not the person.

How is the AI exposure score for National government administrative occupations calculated?

Each official task statement for the occupation is rated on five published 0–4 dimensions (output replicability, physical embodiment, licensed accountability, real-time human trust, and data availability) by claude-opus-5 using scoring prompt task_scoring_v1.0. The model never writes the score; a published formula turns the five ratings into a 0–100 number, so every score can be recomputed by hand. The occupation figure is the importance-weighted mean across 55 scored tasks. The prompt, the rubric, the formula and the full dataset are published at https://futureproof.collab365.com/method and https://futureproof.collab365.com/data/2026-q4.1 under CC BY 4.0.

UK dataswitch to US

National government administrative occupations

calculating taxes owed and ensuring compliance with payment, completing vat returns and issuing a refund for overpayment. If that's your week, this page is about your job.

The honest answer

This job is splitting in two: checking tax forms for correct names and National Insurance numbers is work AI now does quickly and cheaply, and conducting independent field audits of income tax returns to verify information is work it can't touch.

Your move: what you can actually do about this ↓

Where the work sits, by task weight

shifting to AI: 51%
changing shape: 22%
staying human: 27%

These bars are tasks changing hands, not people being counted out. The ledger below shows which.

Whole-job exposure score 57 out of 100 (52–63 allowing for uncertainty): partial exposure, across 55 scored tasks. The number is the support for the sentence above it, not a headline about anyone’s future.

How we know this

What is measured: Every published task statement for national government administrative occupations is rated on five dimensions: can a model produce the output, does the work need a body in a room, does it need a legally accountable person, does it depend on a person being trusted in the moment, and how much data exists. A published formula turns those five ratings into the score; the model never writes the number.

How the bar is built: Each task’s share of the bar is its published importance weight, so a task you do all day counts for more than one you do twice a year.

Release: 2026-q4.1, scores computed 2026-08-04. Read the full method.

Which half fills your week decides your exposure. That is more in your control than it sounds.

Your week, as this page understands it

National government administrative occupations undertake a variety of administrative and clerical duties in national government departments, and in local offices of national government departments. The job title says “national government administrative occupations”. The real job is the part underneath: conducting independent field audits of income tax returns to verify information. That is the thing someone has to be right about.

The exposed part of this job is specific, and we won’t pretend it is coming back. But national government administrative occupations is not one task. It is 55 scored ones, and they are nowhere near equally exposed. The most durable of them, on this evidence, is conducting independent field audits of income tax returns to verify information, and the ledger below shows exactly why.

Where the work sits, by task weight

shifting to AI: 51%
changing shape: 22%
staying human: 27%

These bars are tasks changing hands, not people being counted out. The ledger below shows which.

How we know this

How the bar is built: Each task’s share of the bar is its published importance weight, so a task you do all day counts for more than one you do twice a year.

Release: 2026-q4.1, scores computed 2026-08-04. Read the full method.

Your job, task by task

These are the official task statements for this occupation, in plain English, sorted by what the evidence says is happening to each one. The official wording sits under every line so you can check the rewrite against it.

2 of this occupation's scored task statements carry a score that was measured under a different occupation's context, because the statement is shared between jobs and has only been scored once so far. Each one names the occupation it was measured under in the free routes below; none is presented as a measurement for this job.
Task clusters are not derived in this release, so the task-cluster field is empty and no Collab365 Space routing is attached to this occupation yet.

Shifting to AI

28 tasks

Tasks today’s tools can already do most of. This is the part we will not soften: where these rows are the bulk of your week, the week changes.

Calculating taxes owed and ensuring compliance with payment
This is reading one thing and writing another: taxes in, a record out. That is the shape today's tools are built for.
importance 90 · 4111/02
Source: “Calculate taxes owed and ensure compliance with payment, reporting, or other tax requirements.” (UK task statement)
How this row was scored
Exposure score: 69 out of 100 (62–76 allowing for uncertainty): high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: Tax calculations follow published rules and software does them accurately, but the resulting demand is an official act of the department.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Answering enquiries about application procedure
This is reading one thing and writing another: enquiries in, a record out. That is the shape today's tools are built for.
importance 90 · 4111/01
Source: “Answer enquiries about application procedure.” (UK task statement)
How this row was scored
Exposure score: 64 out of 100 (57–71 allowing for uncertainty): high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Application procedures are written down, so a system can answer most enquiries, though callers sometimes need a person.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Checking tax forms for correct names and National Insurance numbers
This is reading one thing and writing another: tax forms in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/02
Source: “Check tax forms for correct names and National Insurance numbers.” (UK task statement)
How this row was scored
Exposure score: 93 out of 100 (89–97 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Matching names and National Insurance numbers against records is straightforward automated checking.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Maintaining knowledge of tax code changes
This is reading one thing and writing another: knowledge of tax code changes in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/02
Source: “Maintain knowledge of tax code changes, and of accounting procedures and theory to properly evaluate financial information.” (UK task statement)
How this row was scored
Exposure score: 83 out of 100 (76–90 allowing for uncertainty): very high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Tax law changes are published in full, so keeping up to date is well suited to automatic summarising.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 0/4 · how much data exists 4/4.

Changing shape

12 tasks

Tasks where the machine takes the producing and a person keeps the checking, the deciding, or the answering-for-it. For most jobs this is the biggest group, and it is where "transformation, not termination" is literally visible.

Collecting taxes from individuals or businesses according to prescribed laws and regulations
The software now makes the first pass at taxes, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 90 · 4111/02
Source: “Collect taxes from individuals or businesses according to prescribed laws and regulations.” (UK task statement)
How this row was scored
Exposure score: 48 out of 100 (41–55 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: Tax collection follows published rules and runs largely through systems, but it is an official act carried out under statutory powers.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Initiating procedures to grant, modify, deny or terminating assistance
The software now makes the first pass at procedures, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 85 · 4111/01
Source: “Initiate procedures to grant, modify, deny, or terminate assistance.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: Software can prepare the decision, but granting or stopping someone's money is a step an accountable officer takes.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Reviewing filed tax returns to determine whether claimed tax reliefs and deductions are allowed by law
The software now makes the first pass at filed tax returns, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 85 · 4111/02
Source: “Review filed tax returns to determine whether claimed tax reliefs and deductions are allowed by law.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: Deciding whether a claimed relief is allowed in law is a judgement an accountable officer signs.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.

Staying human

15 tasks

Tasks that stay with a person, because they happen in the physical world, because the rules need someone accountable, or because the value is that a specific person does them.

Interviewing and investigating applicants for public support to gather information pertinent to their applications
The value here is that a specific person handles applicants and stands behind it. That is earned, not computed.
importance 85 · 4111/01
Source: “Interview and investigate applicants for public support to gather information pertinent to their applications.” (UK task statement)
How this row was scored
Exposure score: 23 out of 100 (16–30 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: the value is that a specific person does it.
The rating behind it: Getting the full picture from a claimant usually needs a real conversation, often face to face.
The five ratings: output a model can produce 2/4 · needs a body in a room 1/4 · needs an accountable person 1/4 · needs to be trusted in the moment 2/4 · how much data exists 2/4.
Conducting independent field audits of income tax returns to verify information
This work happens in the physical world: independent field audits of income tax returns, in a real place. Software cannot follow it there.
importance 85 · 4111/02
Source: “Conduct independent field audits of income tax returns to verify information.” (UK task statement)
How this row was scored
Exposure score: 16 out of 100 (9–23 allowing for uncertainty): minimal exposure, medium confidence.
Why it sits in this group: work that happens in the physical world; someone qualified has to answer for it.
The rating behind it: A field audit means going to the business and testing what is really there.
The five ratings: output a model can produce 2/4 · needs a body in a room 2/4 · needs an accountable person 2/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Assisting applicants in completing application forms for job referrals or Jobseeker's Allowance claims
The value here is that a specific person handles applicants and stands behind it. That is earned, not computed.
importance 80 · 4111/01
Source: “Assist applicants in completing application forms for job referrals or Jobseeker's Allowance claims.” (UK task statement)
How this row was scored
Exposure score: 39 out of 100 (32–46 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; the value is that a specific person does it.
The rating behind it: Forms can be filled in with digital help, but many claimants need someone sitting with them.
The five ratings: output a model can produce 3/4 · needs a body in a room 1/4 · needs an accountable person 0/4 · needs to be trusted in the moment 2/4 · how much data exists 3/4.

Show the other 45 tasks

Inputting tax return information into computers for processing
shifting to AI
This is reading one thing and writing another: tax return information in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/02
Source: “Input tax return information into computers for processing.” (UK task statement)
How this row was scored
Exposure score: 93 out of 100 (89–97 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Keying return information into the system is data entry that software handles directly.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Monitoring loans and accounts to ensure payments are up to date
shifting to AI
This is reading one thing and writing another: loans in, a record out. That is the shape today's tools are built for.
importance 65 · 4111/02
Source: “Monitor loans and accounts to ensure payments are up to date.” (UK task statement)
How this row was scored
Exposure score: 93 out of 100 (89–97 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Monitoring whether accounts are up to date is structured tracking departmental systems already do automatically.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Verifying computations and amounts against supporting documentation
shifting to AI
This is reading one thing and writing another: computations in, a record out. That is the shape today's tools are built for.
importance 75 · 4111/02
Source: “Verify computations and amounts against supporting documentation.” (UK task statement)
How this row was scored
Exposure score: 88 out of 100 (84–92 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Checking sums against the paperwork behind them is exactly what accounting software already does.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 4/4.
Monitoring the payments of benefits throughout the duration of a claim
shifting to AI
This is reading one thing and writing another: the payments of benefits throughout the duration in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/01
Source: “Monitor the payments of benefits throughout the duration of a claim.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Watching benefit payments across a claim is structured monitoring that departmental systems already flag automatically.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Reviewing records for completeness, accuracy and compliance with regulations
shifting to AI
This is reading one thing and writing another: records in, a record out. That is the shape today's tools are built for.
importance 75 · 4111/01
Source: “Review records for completeness, accuracy, and compliance with regulations.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Checking records are complete and compliant is rule-based reviewing software does reliably, feeding a decision taken elsewhere.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Preparing business or government forms for submission
shifting to AI
This is reading one thing and writing another: business in, a record out. That is the shape today's tools are built for.
importance 70 · 4111/01
Source: “Prepare business or government forms for submission.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Filling in standard business and government forms is highly structured work AI does reliably before submission.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Writing reports and keeping records on legal consultations
shifting to AI
This is reading one thing and writing another: reports in, a record out. That is the shape today's tools are built for.
importance 65 · 4111/01
Source: “Write reports and keep records on legal consultations.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Writing up consultations and keeping the file straight is routine documentation.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Preparing forms to facilitate access to medical services
shifting to AI
This is reading one thing and writing another: forms in, a record out. That is the shape today's tools are built for.
importance 60 · 4111/01
Source: “Prepare forms to facilitate access to medical services.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Filling in standard forms so someone can access medical services is templated paperwork.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Processing individual and corporate income tax returns
shifting to AI
This is reading one thing and writing another: individual in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/02
Source: “Process individual and corporate income tax returns.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Working returns through the system is high-volume processing that software already carries.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Sending notices to taxpayers when accounts are in arrears
shifting to AI
This is reading one thing and writing another: notices in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/02
Source: “Send notices to taxpayers when accounts are in arrears.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Arrears notices are triggered by the account balance and follow a fixed wording.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Requesting that the HM Revenue & Customs prepare a return on a taxpayer's behalf in cases where taxes have not
shifting to AI
This is reading one thing and writing another: the HM revenue & customs prepare a return in, a record out. That is the shape today's tools are built for.
importance 70 · 4111/02
Source: “Request that the HM Revenue & Customs prepare a return on a taxpayer's behalf in cases where taxes have not been filed.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Asking HM Revenue & Customs to file on a taxpayer's behalf follows a set procedure and standard wording.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Preparing correspondence to inform concerned parties of licensing decisions or appeals processes
shifting to AI
This is reading one thing and writing another: correspondence in, a record out. That is the shape today's tools are built for.
importance 65 · 4111/03
Source: “Prepare correspondence to inform concerned parties of licensing decisions or appeals processes.” (UK task statement)
How this row was scored
Exposure score: 81 out of 100 (77–85 allowing for uncertainty): very high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Letters explaining a licensing decision or how to appeal are templated correspondence AI drafts accurately from the case record.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Arranging interviews for benefits claimants to address eligibility queries
shifting to AI
This is reading one thing and writing another: interviews in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/01
Source: “Arrange interviews for benefits claimants to address eligibility queries.” (UK task statement)
How this row was scored
Exposure score: 79 out of 100 (75–83 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Booking interviews around diaries and eligibility queries is scheduling work that software already handles well.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Contacting taxpayers by post or telephone to address discrepancies and to request supporting documentation
shifting to AI
This is reading one thing and writing another: taxpayers in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/02
Source: “Contact taxpayers by post or telephone to address discrepancies and to request supporting documentation.” (UK task statement)
How this row was scored
Exposure score: 79 out of 100 (75–83 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Chasing discrepancies and asking for documents is standard correspondence that can be generated.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Informing licence holders about regulations
shifting to AI
This is reading one thing and writing another: licence holders in, a record out. That is the shape today's tools are built for.
importance 70 · 4111/03
Source: “Inform licence holders about regulations related to licensing, permits, and passports.” (UK task statement)
How this row was scored
Exposure score: 79 out of 100 (75–83 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Licensing, permit and passport rules are published, so accurate explanations can be produced automatically.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Maintaining records of contacts, accounts and orders
shifting to AI
This is reading one thing and writing another: records of contacts, accounts and orders in, a record out. That is the shape today's tools are built for.
importance 70 · 4111/02
Source: “Maintain records of contacts, accounts, and orders.” (UK task statement)
How this row was scored
Exposure score: 75 out of 100 (68–82 allowing for uncertainty): high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Keeping contact, account and order records up to date is ordinary database work.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Informing taxpayers of any overpayment or underpayment
shifting to AI
This is reading one thing and writing another: taxpayers of any overpayment in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/02
Source: “Inform taxpayers of any overpayment or underpayment.” (UK task statement)
How this row was scored
Exposure score: 69 out of 100 (65–73 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Telling someone they have paid too much or too little is a standard letter generated from the figures.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Completing vat returns
shifting to AI
This is reading one thing and writing another: vat returns in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/02
Source: “Complete vat returns.” (UK task statement)
How this row was scored
Exposure score: 69 out of 100 (65–73 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: VAT returns are prepared from accounting exports and software does this well, but the return is a formal declaration someone owns.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Issuing a refund for overpayment
shifting to AI
This is reading one thing and writing another: a refund in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/02
Source: “Issue a refund for overpayment.” (UK task statement)
How this row was scored
Exposure score: 69 out of 100 (62–76 allowing for uncertainty): high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: The refund can be calculated and prepared automatically, though releasing public money needs authorisation.
The five ratings: output a model can produce 4/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Evaluating recorded data to determine eligibility status
shifting to AI
This is reading one thing and writing another: recorded data in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/01
Source: “Evaluate recorded data to determine eligibility status.” (UK task statement)
How this row was scored
Exposure score: 66 out of 100 (62–70 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Weighing recorded details against written eligibility rules is desk work that software can draft reliably.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Interpreting information, eligibility requirements, application details and payment methods
shifting to AI
This is reading one thing and writing another: information, eligibility requirements, application details and payment methods in, a record out. That is the shape today's tools are built for.
importance 80 · 4111/01
Source: “Interpret information including eligibility requirements, application details, and payment methods.” (UK task statement)
How this row was scored
Exposure score: 66 out of 100 (62–70 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Reading across eligibility rules, application details and payment methods is documented desk work.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Researching court information to determine the status of liens
shifting to AI
This is reading one thing and writing another: court information in, a record out. That is the shape today's tools are built for.
importance 65 · 4111/02
Source: “Research court information to determine the status of liens, mortgages, or financial statements for tax payment claims.” (UK task statement)
How this row was scored
Exposure score: 66 out of 100 (59–73 allowing for uncertainty): high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Checking court and registry records on charges and mortgages is online search work software does quickly.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Evaluating claim applications, certificates and records to verify information and determine eligibility or validity
shifting to AI
This is reading one thing and writing another: claim applications, certificates and records in, a record out. That is the shape today's tools are built for.
importance 85 · 4111/03
Source: “Evaluate claim applications, certificates, and records to verify information and determine eligibility or validity.” (UK task statement)
How this row was scored
Exposure score: 66 out of 100 (62–70 allowing for uncertainty): high exposure, high confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Comparing applications and certificates against the rules is documented checking work.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Referring applicants to other agencies for assistance
shifting to AI
This is reading one thing and writing another: applicants in, a record out. That is the shape today's tools are built for.
importance 70 · 4111/01
Source: “Refer applicants to other agencies for assistance.” (UK task statement)
How this row was scored
Exposure score: 64 out of 100 (57–71 allowing for uncertainty): high exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Matching someone's situation to the right agency is a look-up task, but local knowledge still matters.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Reporting law or regulation violations to appropriate boards or agencies
changing shape
The software now makes the first pass at law, but the deciding part still needs a person. So the job becomes checking and deciding rather than producing.
importance 70 · 4111/03
Source: “Report law or regulation violations to appropriate boards or agencies.” (UK task statement)
How this row was scored
Exposure score: 59 out of 100 (52–66 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; mistakes that are cheap to catch.
The rating behind it: Reporting a breach to the right body is documented procedure, though deciding it is a breach relies on internal evidence.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 0/4 · how much data exists 2/4.
Explaining applicants' legal rights regarding benefits and grants
changing shape
The software now makes the first pass at applicants' legal rights regarding benefits, but the deciding part still needs a person. So the job becomes checking and deciding rather than producing.
importance 80 · 4111/01
Source: “Explain applicants' legal rights regarding benefits and grants.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Benefit rights are published in detail, so clear written explanations can be produced automatically.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Calculating and authorising amounts of assistance
changing shape
The software now makes the first pass at amounts of assistance, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 75 · 4111/01
Source: “Calculate and authorise amounts of assistance for programmes including grants, monetary payments, and food vouchers.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: The sums can be calculated automatically, but authorising public money out is an accountable person's act.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Providing taxpayers with information and advice for correct tax form completion
changing shape
The software now makes the first pass at taxpayers, but the deciding part still needs a person. So the job becomes checking and deciding rather than producing.
importance 80 · 4111/02
Source: “Provide taxpayers with information and advice for correct tax form completion.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; mistakes that are cheap to catch.
The rating behind it: Guidance on filling in tax forms is fully published, so clear answers can be produced automatically.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Reviewing taxpayer accounts to determine audit requirements
changing shape
The software now makes the first pass at taxpayer accounts, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 75 · 4111/02
Source: “Review taxpayer accounts to determine audit requirements.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: Risk-scoring accounts suits software well, but choosing to open a compliance check is a decision an authorised officer takes.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Submitting contested tax decisions for appeals hearings
changing shape
The software now makes the first pass at contested tax decisions, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 70 · 4111/02
Source: “Submit contested tax decisions for appeals hearings.” (UK task statement)
How this row was scored
Exposure score: 56 out of 100 (49–63 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: Preparing an appeal referral is documented work, but submitting it is an official step.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 3/4.
Providing social workers with pertinent information gathered during applicant interviews
changing shape
The software now makes the first pass at social workers, but the deciding part still needs a person. So the job becomes checking and deciding rather than producing.
importance 65 · 4111/01
Source: “Provide social workers with pertinent information gathered during applicant interviews.” (UK task statement)
How this row was scored
Exposure score: 50 out of 100 (43–57 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; mistakes that are cheap to catch.
The rating behind it: Summarising interview notes for a social worker is drafting work AI does well, though the useful detail comes from the conversation.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Locating assets through third parties to investigate claims of inability to pay taxes
changing shape
The software now makes the first pass at assets, but the deciding part still needs a person. So the job becomes checking and deciding rather than producing.
importance 75 · 4111/02
Source: “Locate assets through third parties to investigate claims of inability to pay taxes.” (UK task statement)
How this row was scored
Exposure score: 50 out of 100 (43–57 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; mistakes that are cheap to catch.
The rating behind it: Tracing assets through third-party records is search work software is good at, though access is restricted.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 1/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Imposing payment deadlines on delinquent taxpayers
changing shape
The software now makes the first pass at payment deadlines, but someone has to be answerable for the result, and it cannot be the software. So the job becomes checking and deciding rather than producing.
importance 80 · 4111/02
Source: “Impose payment deadlines on delinquent taxpayers.” (UK task statement)
How this row was scored
Exposure score: 48 out of 100 (41–55 allowing for uncertainty): partial exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; the same decision, made over and over; someone qualified has to answer for it.
The rating behind it: The deadline letter writes itself, but imposing it carries legal weight an officer takes on.
The five ratings: output a model can produce 3/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 1/4 · how much data exists 3/4.
Inspecting account books and accounting systems
staying human
The rules require a named, qualified person to answer for account books, and that person cannot be a piece of software.
importance 75 · 4111/02
Source: “Inspect account books and accounting systems for efficiency, effectiveness, and use of accepted accounting procedures to record transactions.” (UK task statement)
How this row was scored
Exposure score: 38 out of 100 (31–45 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: reading one thing and writing another; someone qualified has to answer for it.
The rating behind it: Software can test whole ledgers quickly, but judging whether accounting procedures are sound is a call an official has to make.
The five ratings: output a model can produce 3/4 · needs a body in a room 1/4 · needs an accountable person 2/4 · needs to be trusted in the moment 0/4 · how much data exists 2/4.
Interviewing doctors, medical specialists or employers to get additional information
staying human
The value here is that a specific person handles doctors, medical specialists or employers and stands behind it. That is earned, not computed.
importance 70 · 4111/01
Source: “Interview doctors, medical specialists, or employers to get additional information.” (UK task statement)
How this row was scored
Exposure score: 35 out of 100 (28–42 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: the value is that a specific person does it.
The rating behind it: Getting extra detail from a doctor or employer depends on a live conversation, though notes can be drafted.
The five ratings: output a model can produce 2/4 · needs a body in a room 0/4 · needs an accountable person 0/4 · needs to be trusted in the moment 2/4 · how much data exists 2/4.
Resolving problems with tax returns in consultation with taxpayers or their representatives
staying human
The rules require a named, qualified person to answer for problems, and that person cannot be a piece of software.
importance 80 · 4111/02
Source: “Resolve problems with tax returns in consultation with taxpayers or their representatives.” (UK task statement)
How this row was scored
Exposure score: 30 out of 100 (23–37 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: the same decision, made over and over; someone qualified has to answer for it; the value is that a specific person does it.
The rating behind it: Sorting out a return with the taxpayer takes back-and-forth discussion and an official decision.
The five ratings: output a model can produce 2/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 2/4 · how much data exists 3/4.
Investigating potentially fraudulent claims
staying human
The ratings behind this row put fraudulent claims well outside what today's tools can do on their own.
importance 75 · 4111/01
Source: “Investigate potentially fraudulent claims.” (UK task statement)
How this row was scored
Exposure score: 28 out of 100 (21–35 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: mistakes that are cheap to catch.
The rating behind it: Fraud patterns can be flagged automatically, but building a case needs interviews and visits.
The five ratings: output a model can produce 2/4 · needs a body in a room 1/4 · needs an accountable person 1/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Determining appropriate methods of debt settlement
staying human
The rules require a named, qualified person to answer for appropriate methods of debt settlement, and that person cannot be a piece of software.
importance 75 · 4111/02
Source: “Determine appropriate methods of debt settlement, such as offers of compromise, attachment of earnings, or seizure and sale of property.” (UK task statement)
How this row was scored
Exposure score: 27 out of 100 (20–34 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: someone qualified has to answer for it.
The rating behind it: Choosing between a compromise, an earnings attachment or seizure uses legal powers only an officer holds.
The five ratings: output a model can produce 2/4 · needs a body in a room 0/4 · needs an accountable person 3/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Investigating irregularities and non-compliance issues
staying human
The rules require a named, qualified person to answer for irregularities, and that person cannot be a piece of software.
importance 75 · 4111/03
Source: “Investigate irregularities and non-compliance issues.” (UK task statement)
How this row was scored
Exposure score: 24 out of 100 (17–31 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: someone qualified has to answer for it.
The rating behind it: Software can flag oddities, but investigating non-compliance means gathering private evidence and using statutory powers an official must exercise.
The five ratings: output a model can produce 2/4 · needs a body in a room 1/4 · needs an accountable person 2/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Warning persons of rule infractions or violations
staying human
The rules require a named, qualified person to answer for persons of rule infractions, and that person cannot be a piece of software.
importance 60 · 4111/03
Source: “Warn persons of rule infractions or violations.” (UK task statement)
How this row was scored
Exposure score: 24 out of 100 (17–31 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: someone qualified has to answer for it.
The rating behind it: Warning someone about a breach often happens in person and carries official weight.
The five ratings: output a model can produce 2/4 · needs a body in a room 1/4 · needs an accountable person 2/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.
Interviewing benefits recipients at specified intervals to certify their eligibility for continuing benefits
staying human
The rules require a named, qualified person to answer for benefits recipients, and that person cannot be a piece of software.
importance 80 · 4111/01
Source: “Interview benefits recipients at specified intervals to certify their eligibility for continuing benefits.” (UK task statement)
How this row was scored
Exposure score: 20 out of 100 (13–27 allowing for uncertainty): low exposure, medium confidence.
Why it sits in this group: someone qualified has to answer for it; the value is that a specific person does it.
The rating behind it: Re-checking someone's benefit entitlement means a real conversation, often face to face, and the decision that follows is an official one.
The five ratings: output a model can produce 2/4 · needs a body in a room 1/4 · needs an accountable person 2/4 · needs to be trusted in the moment 2/4 · how much data exists 2/4.
Securing a taxpayer's agreement to settle a tax assessment
staying human
The rules require a named, qualified person to answer for a taxpayer's agreement, and that person cannot be a piece of software.
importance 70 · 4111/02
Source: “Secure a taxpayer's agreement to settle a tax assessment.” (UK task statement)
How this row was scored
Exposure score: 13 out of 100 (6–20 allowing for uncertainty): minimal exposure, medium confidence.
Why it sits in this group: someone qualified has to answer for it; the value is that a specific person does it.
The rating behind it: Getting someone to agree a settlement depends on persuasion built in a live conversation.
The five ratings: output a model can produce 1/4 · needs a body in a room 0/4 · needs an accountable person 2/4 · needs to be trusted in the moment 3/4 · how much data exists 2/4.
Representing cases in informal appeals hearings
staying human
The rules require a named, qualified person to answer for cases, and that person cannot be a piece of software.
importance 75 · 4111/02
Source: “Represent cases in informal appeals hearings.” (UK task statement)
How this row was scored
Exposure score: 11 out of 100 (4–18 allowing for uncertainty): minimal exposure, medium confidence.
Why it sits in this group: someone qualified has to answer for it; the value is that a specific person does it.
The rating behind it: Arguing a case at a hearing is done by an officer who can be held to account.
The five ratings: output a model can produce 1/4 · needs a body in a room 1/4 · needs an accountable person 3/4 · needs to be trusted in the moment 2/4 · how much data exists 2/4.
Assisting in searching and seizing records to prepare charges and documentation for court cases
staying human
This work happens in the physical world: searching, in a real place. Software cannot follow it there.
importance 70 · 4111/02
Source: “Assist in searching and seizing records to prepare charges and documentation for court cases.” (UK task statement)
How this row was scored
Exposure score: 10 out of 100 (3–17 allowing for uncertainty): minimal exposure, medium confidence.
Why it sits in this group: work that happens in the physical world; someone qualified has to answer for it.
The rating behind it: Searching premises and seizing records has to be done in person under legal powers.
The five ratings: output a model can produce 1/4 · needs a body in a room 2/4 · needs an accountable person 3/4 · needs to be trusted in the moment 0/4 · how much data exists 2/4.
Serving legal documents to relevant parties in court proceedings
staying human
This work happens in the physical world: legal documents, in a real place. Software cannot follow it there.
importance 65 · 4111/02
Source: “Serve legal documents to relevant parties in court proceedings.” (UK task statement)
How this row was scored
Exposure score: 0 out of 100 (0–4 allowing for uncertainty): minimal exposure, high confidence.
Why it sits in this group: work that happens in the physical world; someone qualified has to answer for it.
The rating behind it: Legal papers have to be physically handed to the right person, which needs someone there.
The five ratings: output a model can produce 0/4 · needs a body in a room 4/4 · needs an accountable person 2/4 · needs to be trusted in the moment 1/4 · how much data exists 2/4.

What this job pays, and how many people do it

Median pay: £33,000a year, before tax, the middle of the range, so half earn more and half earn less.ashe-t14, 2025 · ASHE 2025 provisional (reference April 2025)Provisional, because the ONS revises this figure in the autumn.
How we know this
Source: ashe-t14
Reference period: ASHE 2025 provisional (reference April 2025)
Rounding: Shown to the nearest £100. The exact published figure is in the downloadable dataset. We do not render pounds the survey cannot support.
People doing this job: 228,700in the UK, 2026.nomis-aps · Apr 2025-Mar 2026 (latest APS 12-month period)This headcount comes from a survey, not a census, so treat it as a good estimate rather than an exact count.

What is deliberately not here: a forecast of how many of these jobs exist in ten years. Where an official projection exists for a market we publish it with its vintage; where it does not, we leave the space empty rather than borrow the other country’s number.

Why this is shifting

The reason is boringly specific. Most of what is shifting here is reading one thing and writing another: tax forms in, a record out. The rows above are exactly that shape: checking tax forms for correct names and National Insurance numbers and calculating taxes owed and ensuring compliance with payment. What it cannot do is be answerable: independent field audits of income tax returns need a named person the rules will accept, and software cannot be that person. Which is why this page talks about your tasks changing, not your job ending.

Your move

Over a pint: what I’d tell you if you were my friend

The exposed part of your job is the biggest part, and I am not going to dress that up: checking tax forms for correct names and National Insurance numbers is work today's tools do quickly and cheaply, and that is not coming back.

So, given all that: 51% of this job's task weight sits in rows the software is already learning, 22% in rows that change shape rather than disappear, and 27% in rows it is nowhere near. That is the position, measured across 55 scored tasks. It is not a forecast about you.

What you have that the software does not is conducting independent field audits of income tax returns to verify information, plus the years of knowing when something looks wrong before you can say why. That is the raw material for everything below.

This week: one thing

Sit on the machine's side of the desk. Pick one real piece of tax forms you would normally do yourself, let whatever software you already have take the first pass at it, and then go through what it produced and write down every single thing it got wrong. One evening this week. Do not fix anything yet. Just catch it.

What you end up holding: a written list of the machine’s mistakes, in your handwriting
How long it takes: an evening, or an hour if you pick one job rather than one client

If you can’t run software on your employer’s or your clients’ data: Do the same hunt on paper. Take one printed piece of tax forms, work through it the way you always do, and mark every point where you made a call rather than followed a rule. Same evening, same list, nothing to log into and nobody to ask permission from. That list is the same artifact: it is the judgment written down.

Over the next 90 days

Change one sentence about what you do. Not on a CV. Out loud, to whoever pays you. From “my job is checking tax forms for correct names and National Insurance numbers” to “I check what the software does and tell you what it means.” Your error list from this week is what makes that sentence true instead of a claim, so use it: show it once, to one person, inside the next ninety days. Same skills, priced as judgment rather than as typing.

Over the next 12 months

Walk toward the end of this job that answers for things, and get it recognised. Pick the one part of conducting independent field audits of income tax returns to verify information you are already best at, and spend the year making it formal: a qualification, a named responsibility, a specialism people ask for by name. Price it honestly: that is evenings, it is months rather than weeks, and the seats are competitive because everyone in your position is looking at the same door. Before you pay for anything, spend an hour with National Careers Service. It is free and government-funded, and it is listed below with the rest of the free routes.

The roads out of here, and why I am not sending you down them

I looked at the obvious moves out of this job, and here is what I found.

I checked the 12 nearest UK occupations to national government administrative occupations (nearest by the work that AI is not taking, not by job title), and none of them survived. The closest was local government administrative occupations: only about 32% of its durable work is work you already do. Your own job splits about 51/49: that share of the list sits in the top exposure band and the rest does not. On this evidence the honest move is inside the job rather than out of it. Become the person who owns “collect taxes from individuals or businesses according to prescribed laws and regulations”, and let the exposed end go.

How that was checked: this job was compared against all 412 UK occupations in this release on their official task statements, and the 12 nearest were examined one by one. A move that turns on an industry, an employer or a qualification rather than on the work itself will not show up in a check like that. And this release carries no licence register, so anything you are weighing needs that looked up separately.

3 moves I checked and rejected

These are the obvious-looking jumps. They are here with their reasons rather than quietly dropped, because the ones that fail are worth knowing about. It is one less thing to turn over at night.

Local government administrative occupations
Why it looked obvious: It came up as a near neighbour because one of your tasks is on their list in the same words: “interview and investigate applicants for public support to gather information pertinent to their applications”. Across the whole of both lists that adds up to about 31% of the work in that job the software is not taking.
Why I am not recommending it: It is closer than most, and still not close enough: about 31% of that job's durable work is already yours, against the 35% I want to see before I will call something a route.
Look at that job’s page anyway →
Financial administrative occupations n.e.c.
Why it looked obvious: It came up as a near neighbour because one of your tasks is on their list in the same words: “interview and investigate applicants for public support to gather information pertinent to their applications”. Across the whole of both lists that adds up to about 21% of the work in that job the software is not taking.
Why I am not recommending it: You would be starting most of it from nothing: about 21% of the durable work in that job is work you do today, and the rest you would be learning while the bills carried on. I will not move you off one melting floe onto another: 58% of its own task list already scores in the top exposure band (62/100 in this release), so the same software is eating it. It is a pay cut, in those words: £28,996 against your £33,000, 12.1% less (ASHE Table 14.7a, 2025 provisional (both)). Retraining to earn less is a decision, not advice.
Look at that job’s page anyway →
Public services associate professionals
Why it looked obvious: It came up as a near neighbour because one of your tasks is on their list in the same words: “interview and investigate applicants for public support to gather information pertinent to their applications”. Across the whole of both lists that adds up to about 11% of the work in that job the software is not taking.
Why I am not recommending it: You would be starting most of it from nothing: about 11% of the durable work in that job is work you do today, and the rest you would be learning while the bills carried on.
Look at that job’s page anyway →

What I’d stop worrying about

A friend tells you what not to spend fear on. This is that list.

The headline number you read somewhere
The big “X% of jobs” figures are about the whole economy, not about you. The number that describes your job is on this page: 51% of its task weight, across 55 scored tasks. Every row behind it is printed above with the source. Check ours; ignore theirs.
“It’s too late for me to become something else”
You are not starting from zero, and the page shows why: conducting independent field audits of income tax returns to verify information is work the software cannot do and you already do it. The move above is a repricing of what you know, not a new career. Nobody who has just left college has that.
“I should learn to code”
Almost certainly not. The value in your job is moving toward checking, deciding and answering for the output, not toward writing the software. A weekend of Python will not change your position; the error list from this week will.
The “obvious” next job everyone suggests
I checked the obvious moves and most of them did not survive. The reasons are printed with the routes above, including the pay and the gate. A move that fails on the numbers is worth knowing about so you can stop turning it over at night.

You are reading the United Kingdom figures

The United States splits this work across more than one official group, of which Compliance Officers is the closest. The pay and employment figures are not directly comparable, and we do not average them together.

Switch to the United States page →partial match

The other groups this work is counted across:

Eligibility Interviewers, Government Programs

In US official statistics this job is counted as Compliance Officers and Eligibility Interviewers, Government Programs. Pay is shown separately for each of those groups (medians cannot be averaged together), while the task list and the scores on this page are for this group only.

Your route through this

Two honest options, and no deadline on either

Free, and complete

The moves above cost nothing. These are the real services that go with them: public, government-funded, and free at the point of use. Nothing on this page is behind an email address or a payment.

In England:
National Careers Service
A free hour with a government-funded careers adviser is worth more than another evening of reading. In England that's the National Careers Service.
Free, government-funded
- Discover your skills and careers (the skills assessment)
In England:
National Careers Service - Find a course
Search what's actually running near you before you spend anything.
Free to search; individual courses vary
In England:
Free courses for jobs (Level 3 qualifications)
A free Level 3 qualification you already qualify for beats a paid course you don't need.
Free for eligible adults
In England:
Skills Bootcamps
Free, up to 16 weeks, and you get a job interview at the end. In England these are Skills Bootcamps - search what's running near you.
Free for eligible adults in England
- Skills Bootcamps - what they are and who can do them (DfE, Skills for Careers)
- Skills Bootcamps - JobHelp (DWP)
In Scotland:
My World of Work (Skills Development Scotland)
In Scotland it's My World of Work, from Skills Development Scotland.
Free, publicly funded
- Skills Development Scotland
In Wales:
Careers Wales
In Wales it's Careers Wales.
Free, Welsh Government-funded
In Northern Ireland:
Careers Service Northern Ireland
In Northern Ireland it's the Careers Service on nidirect.
Free, Department for the Economy-funded

A nearby route

There's no Space built for civil service administrators yet.

Collab365 Spaces is built by the same people as this site. We find the problems that AI and automation are creating inside one kind of work, then solve them as short courses, briefings and Blueprints. Each Space is the community too, so the research and the people doing your job are in the same place.

What a Space actually is, in full

The closest match is Microsoft 365 Productivity Workers, a community for coordinators, administrators and team leads who run their working day in Outlook, Teams, meetings, files and task lists. It overlaps with the part of your job that is growing: the coordinating half of the job - casework queues, files, follow-ups and meeting actions - in the tools the department already pays for. If that overlap isn't you, the free route below covers the same ground.

Try Microsoft 365 Productivity Workers free →

7 days free, no card needed. Explore up to 2 Spaces before you choose a plan: you pick a plan later, not now.

After the trial it is a paid community, and you get identical data either way. If the overlap above is not your job, the moves above cost nothing and stand on their own.

Noted, and thank you. We’ll email you if a Space for civil service administrators launches. Nothing else.

That did not look like an email address, so nothing was saved. Have another go below.

We could not save that. The fault is ours, not yours, and your address was not stored. Please try again later.

No deadline on any of this. The page will still be here, and the data is refreshed on a published schedule rather than when someone wants a headline.

Questions people ask about this job

Will AI replace National government administrative occupations?: Not as a job, but it is already doing parts of the work. Across the 55 official task statements scored for National government administrative occupations (United Kingdom, SOC 4111), 51% of the importance-weighted core work is made of tasks today's AI could already do most of. The overall exposure score is 57 out of 100 (range 52–63, band: partial). That is a statement about tasks, not about headcount: this measures what AI could do, not whether any employer adopts it, whether the law allows it, or whether doing the routine parts faster creates more demand for the human parts. Figures are from release 2026-q4.1.
Which tasks in “National government administrative occupations” can AI already do?: The highest-scoring tasks in release 2026-q4.1 are: “Check tax forms for correct names and National Insurance numbers” (93/100, very high); “Input tax return information into computers for processing” (93/100, very high); “Monitor loans and accounts to ensure payments are up to date” (93/100, very high). Each score comes from five published 0–4 ratings turned into a number by a published formula, and each carries the model's one-sentence reason on the page.
Which tasks in “National government administrative occupations” stay human?: About 27% of this job's task weight sits in work that scores low for AI exposure. The lowest-scoring tasks in release 2026-q4.1 are: “Serve legal documents to relevant parties in court proceedings” (0/100, minimal); “Assist in searching and seizing records to prepare charges and documentation for court cases” (10/100, minimal); “Represent cases in informal appeals hearings” (11/100, minimal). Low scores usually mean the task needs a body in a room, a legally accountable human, or trust built in real time. Those are the three things the scoring rubric treats as gates rather than obstacles.
What should someone working in “National government administrative occupations” do about AI?: Start from the ledger rather than the headline: 51% of this job's weighted core work is exposed, and roughly 27% is not. The practical move is to spend more of your week on the tasks that score low, the ones above, and to get fluent at directing AI through the tasks that score high, because those are the parts that change whether or not you are ready for them. This page does not predict your job, and nothing here is career advice tailored to you: the score describes the occupation, not the person.
How is the AI exposure score for National government administrative occupations calculated?: Each official task statement for the occupation is rated on five published 0–4 dimensions (output replicability, physical embodiment, licensed accountability, real-time human trust, and data availability) by claude-opus-5 using scoring prompt task_scoring_v1.0. The model never writes the score; a published formula turns the five ratings into a 0–100 number, so every score can be recomputed by hand. The occupation figure is the importance-weighted mean across 55 scored tasks. The prompt, the rubric, the formula and the full dataset are published at https://futureproof.collab365.com/method and https://futureproof.collab365.com/data/2026-q4.1 under CC BY 4.0.

Where these numbers come from

About the data on this page

Provisional, because the ONS revises this figure in the autumn.
This headcount comes from a survey, not a census, so treat it as a good estimate rather than an exact count.
The match between this job and its counterpart in the other country is partial, so the two pages count slightly different groups of people.
2 of this occupation's scored task statements carry a score that was measured under a different occupation's context, because the statement is shared between jobs and has only been scored once so far. Each one names the occupation it was measured under in the free routes below; none is presented as a measurement for this job.
Task clusters are not derived in this release, so the task-cluster field is empty and no Collab365 Space routing is attached to this occupation yet.

Task statements: gaisi-indexProcessing: catalogue-bridge → ssc-relatedness-weighting → task-scoring → score-aggregation
Task weights: gaisi-index (relatedness)
Scores: Rubric task_scoring_v1.0, model claude-opus-5, computed 2026-08-05.
Pay and employment: ashe-t14 (ASHE 2025 provisional (reference April 2025))nomis-aps (Apr 2025-Mar 2026 (latest APS 12-month period))

Figures on this page come from release 2026-q4.1, published 2026-08-05. Every release keeps its own permanent address, so a figure you cite in March is still there, unchanged, in November.

The plain-English wording on this page is assembled directly from the task statements and the published ratings, not written by hand for this occupation. That is why it is specific, and it is also why we say so.

The routes and free resources further up are today’s, not the release’s (last reviewed 2026-08-05). A route is an offer, not a historical fact, so it moves on its own clock.

How we score a job Download this release Look up another job

Using these figures?

Cite this

Everything on this site is published under CC BY 4.0. Quote it, chart it, sell something built on it. Just say where it came from, and cite the dated release rather than the site, so the figure you quote stays checkable.

Plain text

Collab365 (2026). Collab365 Futureproof: task-level AI exposure for US and UK occupations, release 2026-q4.1 (methodVersion 2.0.0, promptVersion task_scoring_v1.0). https://futureproof.collab365.com/data/2026-q4.1. Licensed CC BY 4.0. Built with O*NET data (USDOL/ETA, CC BY 4.0); ONS data (Open Government Licence v3.0); GAISI task framework (arXiv:2507.22748, MIT); BLS data (public domain).

BibTeX

@misc{collab365futureproof2026q41,
  title        = {Collab365 Futureproof: task-level AI exposure for US and UK occupations, release 2026-q4.1},
  author       = {{Collab365}},
  year         = {2026},
  url          = {https://futureproof.collab365.com/data/2026-q4.1},
  note         = {Release 2026-q4.1, methodVersion 2.0.0, promptVersion task_scoring_v1.0, CC BY 4.0}
}

Permalink

https://futureproof.collab365.com/data/2026-q4.1

Data as of release 2026-q4.1, published 2026-08-05. Releases never change after publication; when the figures move, a new dated release is published beside this one and this one stays exactly where it is.