•
HKUST found 72.7% of Hong Kong employees use AI weekly versus 31% globally. Most were never trained; most stay on ChatGPT-style drafting, not data analysis. The practical barrier is time. Here is what one after-work Python sitting actually is.
Hong Kong is not waiting to adopt AI at work. It already has.
An HKUST workplace AI study published on 10 August 2026 asked 3,722 employed people in March and April 2026. More than 60% were aged 30–49 — the core of this city’s working week, not a student sample. 72.7% said they use AI at work daily or weekly. The study compared that with a global figure of 31% (KPMG, 2025). SCMP’s same-day report put the headline in English: more than double the world average, among people who already have a job.
That is the first fact. The second is less flattering. More than 70% wanted AI-related skills. Nearly 60% had never received any related training. You can use a chatbot every Thursday and still have nobody ever sat you down to teach you how the thing actually works.
This article is about that gap — not about which classroom is bigger. If you are choosing a format (large literacy class versus short hands-on sittings), we already wrote After-work Python in Hong Kong: large AI classes vs short hands-on lessons. This piece stays with the survey: what people already do, what they still cannot do, and why “I have no time” is the constraint that keeps the gap open.
Public literacy classes exist in Hong Kong; they are not the subject here.
Among the 3,100-plus respondents who already use AI regularly, generative tools dominate. HKUST reported that 81.66% of those frequent users open generative AI daily or weekly. SCMP named the familiar stack: ChatGPT, Copilot, Gemini — mainly drafting documents, summarising information, and supporting creative work.
That is office English. It is the email you polish on the MTR, the briefing you compress before a 10am meeting, the first draft you would rather not start from a blank page. It is useful. It is also a narrow slice of what “AI at work” sounds like in a job advert.
SCMP recorded the other number that matters: only about a quarter of respondents regularly used more advanced applications — data analytics, image or video generation, or agentic AI that can run multi-step tasks with little oversight. The majority of weekly users are not analysing a table. They are prompting a chatbot to write.
If your week looks like that, you are not behind Hong Kong. You are Hong Kong, as this sample measured it: high adoption, shallow practice, almost no training.
HKUST did not publish a ranked list of “barriers to training.” It published a workforce that wants skills and has not been trained. For employed people aged 30–49 — overtime, commute, dinner, a child or a parent, a 9am desk that does not move — the reason a course never starts is rarely a lack of ChatGPT. It is the missing hour.
That is the barrier this article takes seriously. Motivation is already in the data: more than seven in ten said they wanted to get better. The missing piece is a sitting small enough to survive a normal Hong Kong weekday.
A three-hour block after a full day is not three hours of learning. It is the evening you already do not have. People then stay where the survey found them: another prompt, another draft, still unable to open a CSV and trust a number they produced themselves.
The next skill after “ask the chatbot to rewrite this” is not a new persona. It is being able to read a file, filter the rows that matter, and print a result you can check. That is ordinary Python. It is not a career switch, not a trading income, and not a promise that you will become a data scientist.
Forget a syllabus for a moment. The habit is smaller than a course.
One sitting. One file.
You come home, or you sit down after the last meeting. You open the same folder you used last time. You open one .py file. You do one small thing — install nothing extra if you can avoid it, change five lines if you cannot. You run it. You see a result. You stop.
That sitting is twenty to forty minutes on a good night, not a lecture that starts at 7pm and ends when you are empty. If the file errors, you fix the error or you leave a one-line note for Thursday. You do not “catch up” by stacking four lessons on Saturday. That only recreates the evening you could not protect.
What the sitting produces is an artefact you can rerun: a script, not a chat transcript. ChatGPT drafting disappears into the next prompt. A file on disk is still there when next month’s export arrives from work.
That is the whole difference between “I use AI weekly” as the survey measured it and “I can do something with a table.” You do not need agentic AI for the second one. You need a language, an interpreter, and enough syntax to change a line without waiting for someone else to press Run.
A first realistic sequence looks like this:
print — Your first Python program that uses a library (not another print).Those are Information Hub essays, not extra products. They exist so the first sitting has a place to land.
The survey gap is real. Closing it does not require a bootcamp, a get-rich story, or a claim that intro Python is data science.
Nineteen published lessons will not make you a data-analysis hire. They will not turn ChatGPT drafting into agentic AI. They will not replace the job you will return to at 9am. They can give you a file that runs, and enough syntax that the next sitting is not a restart from zero.
A later path — Python data analysis and machine learning — is an outline we are building. It is not a live 39-lesson product today. Do not enroll expecting that catalogue to already exist.
Stay off unpublished trading or macro material from this page. The survey says most people are still on drafting. Match the work to that fact.
If the habit above is what you need, open Introduction to Python Programming. It currently has 19 published lessons, written to be taken a few evenings a week. Check that course page for the current enrollment window — windows move, and this article will not freeze a date into the title.
Do the first sitting on a night that already exists in your week. The survey already told you the city uses AI. It also told you most people were never trained. The smallest honest response is one file, finished before midnight.