Skip to content
man bot
Compare

Clerical support workers · code 4414

Scribes and related workers

AI can fill in forms. Helping someone who can't read understand a letter still needs a person.

AI could take a fair chunk of all three tasks: writing letters and filling in forms for others, reading documents aloud to people who can't read, and helping with official paperwork.

What they actually do (and what else they get called)

Scribes and related workers write letters and complete forms on behalf of persons who are unable to read or write. Tasks include - (a) reading letters and other written matter to persons who are unable to read or write and providing necessary interpretation and information; (b) writing letters and completing forms on behalf of others; (c) offering advice to individuals and interpreting and helping with the completion of government and other official forms. Examples of the occupations classified here: - Scribe

42 / 100

Changing significantly

AI could do a real share of the everyday tasks. The job is changing shape, not vanishing.

Not the odds you keep your job. It's how much of the everyday work still needs a person, out of 100.

← Robot's turf Human's turf →

The ILO's official verdict: Exposed: Gradient 3

413 of 427 for holding up against AI

Why this score?

The ILO rated each of this job's 3 tasks for how much of it AI could do. They average 58.3 out of 100, so it holds up at 100 − 58.3 = 41.7.

That's the whole sum. No black box, no secret sauce: you can check it against the ILO's study. ILO group “Exposed: Gradient 3”: Most of the tasks are affected.

This job rates 55 out of 100 for how hands-on it is. The score above only looks at AI tools, not robots, and for work like this robots are the bigger question.

from the source Working Paper 140, Generative AI and Jobs 2025 index data 0.8-observed-use

Man v Bot · compare

Who holds up better?

Scribes and related workers 42

v

Go on, try your partner's job. Or your sibling's. We won't tell.

Man v Bot

Where the bot helps, and where the human wins

Bot

What AI takes

The routine stuff. You may not miss it.

  • Writing letters.
  • Filling in forms.
  • Reading documents.

Man

What people still do better

Judgement, trust and skilled hands. Hard to download.

  • Trust and help in person.
  • Explaining what a letter means.
  • Advice on official matters.

Both

What's arriving

New work that only exists because the AI does.

  • AI reading and translation apps.
  • Digital inclusion help.
  • Advice services.

Written by a human, from the ILO's task ratings below. Yes, we see the irony. Read it as one person's take.

Meet your two rivals

They're not the same machine. AI software can only reach the parts of a job done on a screen or in words. A chatbot can't climb a ladder. A robot might, one day, if it doesn't fall off.

The AI

Lives in a screen. Never had to fix a boiler.

Threat level: High

68 could do without you

It could take over a lot of this job on its own. It could help with 75 of 100.

The robot

The wildcard. Has hands, still working out stairs.

How much it matters here: Some

55 how hands-on

A fair part of this job is hands-on, so robots would be needed for that part, not just software.

So could a robot do the physical part?

Physical work where helping and caring for people is a big part of the job. For care work, whether a machine could do the movements is only half the question. The other half is whether anyone wants one to. The US data counts helping workmates too, so a few trades end up here.

Robots hate fiddly work in awkward spaces. On that measure, this job scores 28 out of 100. The higher it is, the longer the robot waits.

55 How
hands-on
28 How fiddly
and cramped

Kind of hands-on work Hands-on, helping and caring for others

The small print

They aren't a robot score, and nothing on this page is. Nobody measures how far robots have got well enough to put a number on it. These are the closest thing we have: how hands-on the job is, and how much fiddly, cramped work it takes, which is what Frey and Osborne found hardest for machines.

From O*NET, averaged across 4 US jobs .

finger skill 28 · hand skill 26.8 · cramped spaces 29.1 · averaged · rule: hands-on 50+ · caring 4+ of 7

estimated O*NET 31.0 BLS occupational crosswalk chain ISCO-08 to 2010 SOC (Aug 2012, rev. Jun 2015); 2010 to 2018 SOC (Nov 2017); O*NET-SOC 2019 taxonomy data 0.8-observed-use measures chosen by Frey & Osborne (2017) · the grouping is a sorting, not a measurement

The job, task by task

Every job is really a bundle of tasks. Here's this one unpacked, with how much of each AI could do, out of 100. The score is the average, and averages can hide a lot.

AI's favourite task here scores 67, against an average of 58. The bigger the gap, the more lopsided the job: some of it wide open to AI, some of it barely touched.

  1. 67

    Writing letters and completing forms on behalf of others

    Medium
  2. 56

    Reading letters and other written matter to persons who are unable to read or write and providing necessary interpretation and information

    Medium
  3. 52

    Offering advice to individuals and interpreting and helping with the completion of government and other official forms.

    Medium

3 tasks · the High to Very Low labels are the ILO's

from the source Working Paper 140, Generative AI and Jobs 2025 index data 0.8-observed-use

Sidekick or replacement?

A tool that makes you faster is a very different thing from one that does the task while you watch. Jobs and Skills Australia measured the two separately, which is handy.

AI helps you
75
AI does it for you
68

Sidekick wins, by 7 points.

How good a match is this?

One Australian job matches this one closely, so this is about as reliable as it gets here.

estimated Our Gen AI Transition (Gen AI Capacity Study) September 2025 detailed data release data 0.8-observed-use matched to Australian jobs 5999

Meanwhile, in real life…

Everything else here is about what AI could do. This is what people actually did. When they asked Claude for help with this kind of work, did they hand the whole task over, or roll their sleeves up and work through it together?

Out of every 100 conversations

46
54

worked through it together handed it over

The small print

Anthropic publishes this about its own product, and it doesn't feed into any score here. These are people who use Claude, not workers in general, so little use here doesn't mean AI can't reach the job. This job makes up 0.1% of the conversations.

The data is sorted by US job, and this one matches 3 of them. We've combined them, counting each by how often it comes up in the conversations, so the busiest one counts most.

These two add up to 100. The bars elsewhere on the page don't, because each is scored on its own, so we don't put them side by side.

observed use Anthropic Economic Index release 2026-06-26, period beginning 2026-05-01 data 0.8-observed-use 2026-05-01 · Anthropic, about its own product

So, should scribes be worried?

Short answer: not worried, but paying attention. The job is changing.

Much of it is exposed. Writing letters, reading for others and helping with forms are all rated Medium. AI can read and write, but people who need this help often need someone they trust.

Worth remembering: we score tasks, not jobs. When AI takes on a task, the job usually reshuffles around it rather than vanishing.

Where this job could lead

None of the new jobs we've written up start from here yet. Could be you're the first. See the jobs AI is creating →

How are the neighbours doing?

There are 29 jobs in the “Clerical support workers” family, averaging 47. This one is below the family average.

Pit this against any other job →