AI isn't replacing people. People who use AI are.

The Drop

AI isn't replacing people. People who use AI are.

Every few months someone asks whether AI will take their job. It's the wrong question. The software doesn't turn up to work on its own. Somebody has to pick the problem, feed it the data and check what comes back. The real split is between people who can do that and people who can't, and this week's news from Vietnam and Cambodia shows it clearly.

Vietnam's National Technology Innovation Foundation (NATIF) is preparing a programme that pays up to half the cost of a company's AI project, capped at VNĐ3 billion (about US$150,000). It's adapted from Singapore's 100 Experiment model, with about 15 projects planned for 2026 and around 50 a year from 2027. The money arrives in stages: 20 per cent once the problem is defined and the data is ready, 30 per cent when the model works, and the last 50 per cent only when it's running in the business and the productivity gain can be measured. Within days of hearing about it, one business association had 26 proposals: 14 companies wanting AI and 12 offering to build it. At the launch meeting, the companies' big question wasn't about the technology. It was how to define a suitable AI problem in the first place.

In Cambodia, the Minister of Posts and Telecommunications said on 23 September that the country has blocked more than 53,000 scam websites and worked with Meta and Telegram to remove more than 53,000 accounts and groups. He also said scams are still mainly run by people, but criminals are using more AI and automation, and the authorities have to keep pace. The scammers have already worked out that a person with AI can do the work of many. The people trying to stop them are the ones catching up.

Opinion. Both stories come down to the same thing. In Vietnam, the state money goes to the manager who can write down one clear problem. In Cambodia, the scam networks got ahead by picking up the tools first. Nobody in either story is being replaced by a machine. They're being outpaced by people who learned the machine first. Joining them is easier than it sounds. You don't need to code. You need one real problem, a few hours with the tools, and the habit of checking what they give back. That's this week's try.

Vietnam News on NATIF -> · Cambodianess on the scam crackdown ->

Agentic Frontier

Funding Tracker

Vietnam will pay up to US$150,000 per company AI project
NATIF covers up to 50 per cent of an approved project's cost, capped at VNĐ3 billion, and pays out as the project hits milestones: problem defined, model built, model running with measurable results. Medium and large manufacturers with real AI needs and enough data get priority; smaller tech firms can join as solution providers.
[Source: Vietnam News]

Policy

Cambodia sets licensing rules for crypto firms
In a joint statement, the National Bank of Cambodia and the Securities and Exchange Regulator of Cambodia split oversight of virtual-asset services. Banks need the central bank's prior authorisation; other providers need an SERC licence. Licensed firms become reporting entities under anti-money-laundering rules. Crypto is still not legal tender, and a digital-asset law is being drafted.
[Source: Kampuchea Thmey] · [Khmer Times]

Security

53,000 scam sites blocked, and why that isn't enough
Analyst Seng Vanly says the operators simply open new sites or move to messaging apps; the fix is to follow the money and prosecute the organisers. Cambodia hosted 27 countries, the FBI and Interpol in Phnom Penh on 23 and 24 September to coordinate a cross-border response.
[Source: Cambodianess]

Try This Week

Write your first AI problem statement

Vietnam's new AI fund makes its first payment when a company has defined its problem and prepared its data. That's the step most businesses skip, and it's the one skill that separates people who use AI from people who just open the chat window.

Spend 15 minutes on one task from your own work. You don't need any coding and any AI assistant will do. The point is to learn how to hand a job over clearly, and how to check the result.