In Vietnam, 18.01% of sampled conversations were matched to software development tasks. In Korea that figure is 10.42%.

What the data says about Vietnam

Vietnam’s overall Anthropic Usage Index sits at 0.53, roughly a seventh of Korea’s 3.78. If you stopped there, you would assume conversations about software are proportionally rarer too. They are not. 18.01% of sampled conversations from Vietnam matched software development tasks — against 11.51% globally and 10.42% in Korea. On the thing I care most about, conversations from Vietnam are 72.8% more likely to look like software work than conversations from Korea. Lower overall usage, heavier engineering flavour in the usage that exists.

The whole profile tilts that way. Conversations matched to “Computer and Mathematical” job tasks ran at 29.32%, versus 23.8% globally — a 23% lead. Meanwhile conversations matched to sales-related tasks ran 29% under the global average, and research-and-intelligence conversations came in at 8.11%, about a quarter below the global 10.94%. Document processing, by contrast, ran 22.9% above. Read together, the picture is of usage concentrated on building and producing concrete output rather than on open-ended investigation — conversations that look like someone with a target and a deadline, not someone exploring what the tool can do.

More than half of conversations — 53.13% — were work-related, against a global share of 43.36%. Another 20.38% looked like coursework or learning, also above the global 16.45%, and the education topic overall ran 29.7% above Korea’s share. And the automation side edges augmentation, 51.38% to 48.62%, while globally the two sit evenly split. My read, and it is only a read: these conversations look less like people delegating tasks wholesale and more like people using these tools as production machinery — get the code written, get the document extracted, get the thing shipped. What that says about the individuals behind the conversations, I genuinely cannot tell you, and neither can the data.

One honest caveat: an index of 0.53 across 121 countries (rank 84) means the sample this snapshot is drawn from is thinner than Korea’s. Strong signals in a smaller base deserve some humility. But the direction is consistent enough — software development over-indexed against the world, against Korea, and against every other topic except content creation — that I would not dismiss it as noise.

This is exactly the territory where I can be useful. If you are in Vietnam and mid-build — wrestling an app through store review, wiring up AI agents or MCP tooling, sorting out a backend or an auth flow, fighting a mobile bug, or hardening something that keeps falling over in production — that is the work I do, and I am happy to look at it with you. Vietnam sits close enough to Seoul’s time zone that finding a slot to talk is about as easy as scheduling gets.

The numbers

VietnamGlobal
Anthropic Usage Index0.53 (rank 84 of 121)1.00 baseline
Work conversations53.13%43.36%
Learning / coursework20.38%16.45%
Automation-leaning51.38%48.62%
Augmentation-leaning48.62%51.38%

Most common request topics

TopicVietnamGlobalDifference
Content Creation & Copywriting24.91%22.72%+10%
Software Development18.01%11.51%+56%
Education & Learning14.54%13.23%+10%
Research & Intelligence8.11%10.94%-26%
Document Processing & Extraction5.31%4.32%+23%

Job-task categories these conversations matched

CategoryShare
Computer and Mathematical29.32%
Arts, Design, Entertainment, Sports, and Media15.04%
Educational Instruction and Library11.02%
Office and Administrative Support7.47%
Sales and Related6.46%

Where Vietnam is ahead of Korea

  • Software Development — 18.01% here, 10.42% in Korea (72.8% higher)
  • Education & Learning — 14.54% here, 11.21% in Korea (29.7% higher)

Ask me something

I’m Seunghan, an engineer in Seoul. I answer questions about shipping apps, AI agent tooling, backends, and reliability — free, and for anyone. Here’s what I can help with and why.

GitHub · KakaoTalk open chat · All countries


Figures from the Anthropic Economic Index, snapshot 2026-05-01. They describe observed Claude conversations matched to job tasks — not who users are, not employment, not the labour market. A country absent from the index means data was not published, never a measured zero.