
Saving Money and AI Tokens: Use AI to Build Tools Instead of Doing the Work
Vũ Khôi
Tác giả tại Wotech Electronics Blog

Saving Money and AI Tokens: Use AI to Build Tools Instead of Doing the Work
Phạm Đăng Định shared how he saves money and AI token quota for his company: 264,000 VND/month in total for AI — ChatGPT Go at 132,000 VND/month for generating images and GG AI Plus at 132,000 VND/month for creating content, analyzing data, and vibe-coding extensions — while covering almost every task. The post presents his 4-part method, with a detailed guide to creating a free Chrome Extension via Google AI Studio. This is the author's personal experience — he says he doesn't know code and has never learned to code; the cost figures and interpretations in this article belong to the author and have not been independently verified by Wotech.
Quick answer: The author's method has 4 parts. Part 1: each week spend about 30 minutes listing all daily tasks — for his SEO content and website administration work: collecting partner website data into Google Sheets, posting approved SEO articles/products to websites (WordPress, Haravan, Sapo), reading and filtering notification emails each morning, checking keyword rankings and index status. Part 2: classify into 2 groups — thinking tasks (keyword analysis, outlining, creative writing, summarizing customer feedback) should use AI; mechanical, purely technical tasks (copying data from web to Sheet, clicking save, parsing HTML, batch renaming files) should not use AI line by line, because they only move information from point A to point B, and using AI to read each article is both slow and a waste of tokens. Part 3: use AI to build tools instead of doing the work — ask AI to write a Chrome Extension, Google Apps Script, or Bookmarklet; his WP-to-Sheet AI Crawler extension automatically collects hundreds of products with 0 AI tokens. Part 4: set up an independent automated operating environment — a separate Chrome Profile, a machine left running in the background, and results checked on Google Sheets at the end of the day.
Vibe coding means asking AI to write code from a natural-language description instead of writing code yourself — the user describes the need, AI generates the full source code, and the user only installs and runs it.
Parts 1 and 2: List, then classify tasks
Each week, the author spends about 30 minutes writing down all the tasks he does every day. For his SEO content and website administration work, the list includes: collecting data from partner websites (product name, price, short description, detailed HTML content) into Google Sheets; posting approved SEO articles or products to websites; reading and filtering new notification emails each morning to make a to-do list; checking keyword rankings or the index status of URLs.
The decisive step for saving tokens, according to him, is clearly distinguishing tasks that need thinking from tasks that are only repetitive:
| Group | Characteristics | Examples | Handling |
|---|---|---|---|
| Group 1: needs thinking | Needs analysis, creativity, judgment | Keyword analysis, outlining, creative writing, summarizing customer feedback | Use AI |
| Group 2: mechanical, purely technical | Only moves information from point A to point B | Copying data from web to Sheet, clicking save, parsing HTML structure, batch renaming files | Don't use AI directly line by line |
The author stresses: for group 2 work, using AI to read each article is both slow and a useless waste of tokens.
Part 3: Use AI to build tools
Instead of making AI read and copy 500 articles, the author asks AI to write a small tool to do it. Depending on the need, you can create a Chrome Extension, Google Apps Script, or Bookmarklet. His detailed guide for creating a free Chrome Extension via Google AI Studio:
Step 1: Go to Google AI Studio, sign in with a normal Google account (the author left the link in the original post's comments).
Step 2: Choose a Gemini model (the Flash or Pro version available with a very large free daily quota and long context, so writing a full extension's code is very smooth, according to the author).
Step 3: Enter a prompt clearly describing the need, in a 4-part structure:
- Role: You are an expert Chrome Extension programmer, Manifest V3 standard.
- Input & processing: Create an extension with a popup UI. When I paste a list of article/product URLs into the input box, the extension browses each link and extracts the fields: Title, HTML Description, Price...
- Output: After extraction, data is sent to Google Apps Script via a Web App URL to write into Google Sheets. Add a customizable delay between links and retry count on network errors.
- File requirements: Write the complete source code for each file: manifest.json, popup.html, popup.js, background.js (no abbreviations, no omitted code).
Step 4: Install the extension in the browser: create a new folder on the computer (e.g. named My-Crawler), save each code file under its correct name in the folder; open Chrome, type chrome://extensions/, enable Developer mode at the top right, click "Load unpacked" and point to the folder just created.
The author's own example — the WP-to-Sheet AI Crawler extension (pictured in the post): enter the list of links to edit into the extension; the extension automatically opens each link in the background and extracts 8 standard data fields (ID, Name, short description, long HTML description, Price, Slug...); automatically pushes the data straight into Google Sheets via a Web App URL, with sequential delay and retry mechanisms to avoid network congestion. The whole process of collecting hundreds of products runs fully automatically and costs 0 AI tokens, according to the author.
Part 4: An independent automated operating environment
So the work doesn't interrupt the computer during main work, the author sets up:
- A separate Chrome Profile (or a separate Google account): only install automation extensions and open data-scraping pages on this profile, avoiding data conflicts and keeping the main personal account safe.
- A machine left running in the background: each morning at work, just paste the list of links to process into the tool, press start, then minimize the window; anyone with an unused old computer can plug it in separately to run all day.
- Check results: at the end of the day or after the tool finishes, just open Google Sheets to verify the data has fully synced.
The author adds: the vibe-coding app he uses is free (information in the comments), this is personal experience in a time when everything is getting more expensive, and he doesn't know code and has never learned to code.
Frequently asked questions
1. How much does the author spend on AI per month? According to the post, 264,000 VND/month in total: ChatGPT Go at 132,000 VND/month for images and GG AI Plus at 132,000 VND/month for content, data analysis, and vibe-coding extensions.
2. What is the task classification principle? According to the post, thinking tasks (analysis, creativity, judgment) should use AI; mechanical, purely technical tasks (only moving information from point A to point B) should not use AI line by line because it is both slow and a useless waste of tokens.
3. What does "use AI to build tools" mean? Instead of making AI read and copy each article, the author asks AI to write a small tool (Chrome Extension, Google Apps Script, or Bookmarklet) to do it — for example, the WP-to-Sheet AI Crawler extension collecting hundreds of products with 0 AI tokens.
4. What are the steps to create a Chrome Extension with Google AI Studio? Per the author's guide: go to Google AI Studio, choose a Gemini model, enter a prompt in the 4-part structure (role, input & processing, output, file requirements), then save the code into a folder and install it via chrome://extensions/ in Developer mode with "Load unpacked".
5. How do automated tools avoid interrupting the main machine? According to the post: create a separate Chrome Profile (or Google account) only for automation, leave a machine running in the background (or use a separate old computer), and check results on Google Sheets at the end of the day.
Source: post by Phạm Đăng Định in a community group (18/11/2026). The cost figures, guides, and interpretations in this article are the author's personal experience and have not been independently verified by Wotech; the author left the Google AI Studio link in the original post's comments.
Bài viết liên quan

18 tháng 11, 2026 · 6 phut doc
FPT Software Chairwoman: US Clients No Longer Want to Pay by the Hour, Company Shifts to Outcome-Based Pricing
Artificial intelligence is forcing FPT Software to transform its traditional software outsourcing model: large clients in the US now apply 100% outcome-based payment, and the company expects the Time…

18 tháng 11, 2026 · 6 phut doc
Meta Expands Muse to Australia: The 6th Country, Real-Time Voice, and Local App Integrations
On 17/11/2026, Meta expanded Muse to Australia — the 6th officially supported country, just over 2 months after its US launch. According to a post by Hồ Hoàng Tiến (18/11/2026), the notable point is …

18 tháng 11, 2026 · 7 phut doc
AI Agent Costs Grow With the Square of Steps: 4 Layers to Cut Token Spend Before Switching to a Cheaper Model
You don't pay for answers — you pay for how many times you re-ask the same thing. That is the central argument of Hoang Viet AI in a post in the comailo group (18/11/2026): an agent running 40 steps,…