An essay has been circulating among Chinese AI builders for months, nicknamed the shortest path to money. Its advice: do not rush to dream about monthly subscription software billed in US dollars. Go one floor down: help a traditional small-business owner digitize documents, wire up data flows, and rebuild processes with AI. The model now has a name: FDE (Forward Deployed Engineer), an engineer deployed on-site at the customer's premises.
The short answer: FDE work in China has become a real profession, with an encyclopedia entry and price lists: an AI engineer sits inside a small business selling packaged engagements, from knowledge-base builds to process restructuring. Vietnam wants the same thing, but prices, local tooling and the legal environment differ enough that the playbook has to be rewritten.
TL;DR
- FDE in China is a real job now: it has an encyclopedia entry, fast-growing hiring, and public price lists for small businesses.
- The economics decide who survives: each delivery must get cheaper, and field experience must become an asset you reuse.
- For Vietnam, keep the three-tier shape but swap the foundations: Zalo and spreadsheets instead of WeChat, smaller deals priced in slices, and a tightening legal regime that rewards disciplined vendors.
On this blog we have covered the Vietnamese SMB market from the guide angle, a complete guide to AI for Vietnamese small and medium businesses, from the sovereignty angle, using agents without falling into vendor lock-in, and from the pricing angle, why transparent agent pricing wins deals. This piece takes a different route: read the FDE essay circulating in China's AI community, line it up against what the market is actually doing, then apply it to Vietnamese numbers.
The essay lays out three very specific price tiers: build a corporate knowledge base for roughly 10,000 RMB (about VND 35 million); connect data across chat tools, spreadsheets and business systems for 30,000 to 50,000 RMB (about VND 105–175 million); then audit each role, replace repetitive work with AI workflows, and upgrade the hard tasks with employee skills. It closes on a line worth remembering: traditional businesses do not buy technology, they buy certainty in their processes, hours they can cut, and core assets that no longer walk out the door with an employee.
We read that essay, checked it against what is really happening in China, then ran it against Vietnamese data. Three things are worth learning. Three things must be rewritten before anyone copies it.
FDE in China is no longer a viral essay — it is a profession with an encyclopedia entry
The biggest difference between that essay and most "make money with AI" content online: it matches what the market is really doing.
Baidu's encyclopedia now carries a dedicated entry for FDE, the Chinese rendering of "forward deployed engineer". Job postings for the role grew more than 800% in the first nine months of 2025. Shanghai's government folded FDE into its AI development plan under a program named "Hundred–Thousand–Ten–Thousand": connect 100 businesses, create 1,000 AI agents, steer 10,000 programmers toward the work, and stock a reserve of more than 1,000 FDE engineers. FDE training was tied to the national mid-level engineer title framework, administered by the human-resources authorities. In October 2026, Shandong followed with a policy of letting application scenarios pull the models and industry data feed them.
The big vendors are louder still. Microsoft committed $2.5 billion to found Frontier Company and put 6,000 specialists into customer environments. AWS is spending $1 billion on its own FDE platform, promising to cut delivery time from months to days. OpenAI, Anthropic and Cohere have all doubled or quintupled their deployment teams. The shared idea: behind every large model contract sits a person physically sitting in the customer's building.
And at the lowest floor, exactly the floor the essay aims at, a layer of FDE workshops serving SMEs has formed with published prices: diagnostics at 4,900 RMB, knowledge-base packages from 9,800 RMB, retainer contracts of 4,000–20,000 RMB a month. The community also circulates a three-tier "FDE technical framework for SMEs": tier one uses off-the-shelf SaaS (1–3 days, ROI measurable to the digit), tier two wires integration platforms such as n8n or Make into Feishu (1–3 weeks), tier three only writes requirements documents for outsourced developers, with no in-house coding.
So where does the essay get it right? It points in the direction the cash flow is already moving. It skips the question that Tencent Cloud's own industry report raises, and that question is the key to Vietnam.
The economics: FDE lives only when delivery costs fall and experience becomes reusable
Tencent Cloud's ecosystem report, "Industry Observation and Practice of the FDE Model", points at a reality nobody enjoys hearing. In Silicon Valley, the big customers of Palantir and OpenAI sign contracts of six or seven figures in USD per year, enough to keep an elite crew embedded on site. Most projects in China sit in the 100,000 to 1 million RMB range. Deliver at elite-consultant rates and you lose money; deliver at outsourced-shop rates and nothing accumulates, and next year you start from zero.
Their conclusion: Chinese FDE shops must solve two problems, drive down the cost of each delivery with AI, and turn field experience into reusable assets: skills, connectors, industry templates. They draw a blunt line: leaving the customer with a system alone is contracting; taking experience home with nothing reusable is a project; converting field experience into reusable assets, with each delivery cheaper than the last, that is FDE.
This is the spot where anyone selling AI services in Vietnam should stop and think. The viral essay paints a picture: help one business, collect VND 35 million. It never says what it costs to produce that VND 35 million, or whether the second engagement takes the same effort.
Vietnam compared: the demand is real, but the shape is different
Vietnamese data from the last two years draws a paradox, and the paradox is exactly the gap the FDE model can slip into.
The Deloitte–Meta report (May 2025): 93% of Vietnamese small and medium businesses have used at least one AI tool, the highest rate among the six markets surveyed and first in Southeast Asia. Very encouraging. But the AWS × Strand Partners report "Unlocking Vietnam's AI Potential" (2026), released at the end of September, offers another slice: of nearly 245,000 businesses using AI, about 61% are still only exploring or experimenting, and just 8% have actually integrated AI into operating workflows. A domestic conference (August 2026) quoted an even starker figure: only about 1% of businesses reach real AI maturity.
In service language: almost everyone has "touched" AI, and almost nobody "lives" on AI. ChatGPT writes captions and slides: that part is done. But the document store still sits on individual laptops, the sales script still travels by word of mouth, a key employee's resignation walks out with the whole playbook, and data still moves by hand between Zalo, Excel and the accounting package. Exactly the chaotic state the Chinese essay describes, only the tools differ.
Three shape differences to price in before you copy anything:
First, the Vietnamese SME's digital shop floor is different. Over there it is WeChat Work, DingTalk and Feishu, ecosystems with APIs that are largely open to bots and automation. Over here it is Zalo, Excel, Google Sheets, MISA, e-invoicing, and a bit of ERP. A Vietnamese FDE has to know where an API exists and where digitization only happens by hand and process. The data-flow tier of the original essay is therefore more labor-intensive, and less contested precisely because it is exhausting.
Second, deal sizes are smaller, and cost is the number-one bottleneck. A UEH survey cited in GIZ's annual report (2025): 86% of SMEs name upfront investment as the main barrier, above even staff shortages. At the same time, 55% of businesses say missing skills are the biggest barrier (figures from the August 2026 conference). In other words, in Vietnam an FDE does not only sell engineering; it sells the pattern of "try small, see payback fast", which is exactly the tiered price-list logic of the Chinese workshops: a cheap diagnostic first, a retainer that nurtures the account after.
Third, the legal door has closed, and that favors vendors who deliver properly. Vietnam now has a dedicated law on artificial intelligence in force since March 2026, alongside a data-protection framework that keeps tightening. SMEs running on shadowy AI, with customer data sent through personal ChatGPT accounts, take on more risk every month. A provider who delivers systems with access controls, data sitting inside the customer's own infrastructure, and traceable audit logs sells something a hobby chatbot cannot compete with.
The three tiers of the Chinese essay, rewritten for the Vietnamese market
Tier one — build the knowledge frame. Still right, still the gentlest way in, and in Vietnam even easier to sell because everyone has lost documents to an ex-employee. But keep one thing clear in your head: this is the hook product. Its real value arrives with the second delivery, once you understand the shape of the customer's data.
Tier two — connect the data flows. In China people wire DingTalk and Feishu. In Vietnam the anchors are Zalo (auto-replies, lead capture, sync into a customer sheet), Google Sheets and MISA for the books, and integration platforms like n8n running on a private server. The original price band of 30,000–50,000 RMB converts to roughly VND 100–175 million. For Vietnamese SMEs that is revenue in the tens of billions per year and up; for everyone else it must be sliced into process-sized packages at lower prices, while templates stay reusable across customers in the same industry.
Tier three — audit each role, restructure the processes. This tier creates the most surplus value and is the one almost nobody executes, because it means sitting down at every desk, listing what each person does daily, and ruling: this becomes an AI workflow, that only needs better skills. The prize sits at the end of this tier: the best closer's playbook, packaged into a shared asset for the whole team, the thing an owner will pay for because it retains capability, not headcount.
What the Chinese essay gets right, and what Vietnamese sellers must check themselves
Check one: 10,000 RMB for a knowledge base is not the Vietnamese market price. Prices here anchor to process outcomes and delivery cost, not to inspiration. Look at our own niche: business agent products like 5ac.vn's G-Company OS start at about $999 in one-time deployment fees (roughly VND 23.5 million) for teams of 5–10 people, below the essay's first tier, with each tier scaling up with team size. The real lesson of that number is not "cheaper than China". It is this: to survive at low prices, your delivery machinery has to be automated to the bone, which is precisely Tencent Cloud's economic lesson.
Check two: "FDE" is a new name for an old trade. Deployment, consulting and outsourcing all existed before, and people died on their margins doing it. The new part is not the letters. It is that for the first time one person plus AI tools does the work of four (advise, build, configure, train). Whoever copies the shell, the title and the business card, without copying the core, reusable assets and falling per-tier cost, will repeat the exact death of the traditional integration trade.
Check three: the original essay's closing line, "businesses buy certainty, not concepts", is the most correct principle in the whole piece, and it holds in any market. It is also the test applied to every AI service seller: if the owner cannot compute money from what you said, you are selling technology, and technology always gets price-compared.
If you have to start today
Pick a business you know. Pick one painful process: quoting, lead follow-up, the weekly report. Estimate what the manual data entry costs in real wages, using the ROI formula Chinese FDE shops use: hours replaced × hourly rate × frequency across a year. Write it on one page, tell the owner exactly that number, and deliver within two weeks, with a set of templates you keep so the next customer runs faster.
Real money is not in software with a million subscribers. It is in how you help a business stop depending on two people.
Frequently asked questions
What is FDE, and how is it different from the deployment staff a business already hires?
FDE stands for Forward Deployed Engineer: a person who sits at the customer's site to digitize documents, connect data, and rebuild processes with AI. Two things separate it from ordinary contract work. Field experience from each engagement must become a reusable asset (skills, templates, connectors), and each delivery must get cheaper thanks to AI. Without both, it is the old integration trade wearing a new name.
Can China's FDE model be applied to Vietnamese SMEs as is?
Right direction, wrong to copy wholesale. Vietnam's digital shop floor runs on Zalo, Excel and MISA instead of WeChat Work, DingTalk and Feishu; deal sizes are smaller; and the AI and data rules are tighter. The three tiers from the original essay survive, but each tier's footing has to be rewritten for the local market.
Where should a solo operator start if they want to sell FDE work in Vietnam?
Start with a cheap diagnostic to get in the door, build the knowledge base as the hook, then earn on tiers two and three, where the real value sits in understanding the customer's data structure. The entry package has to be cheap with fast payback, because upfront cost is the biggest barrier for Vietnamese SMEs, according to the UEH survey in GIZ's annual report.
How does Vietnam's AI law affect someone selling AI services to SMEs?
Vietnam now has a dedicated law on artificial intelligence in force since March 2026, alongside a data-protection framework that keeps tightening. Businesses running AI through personal accounts get riskier by the month; a provider who ships systems with access controls, data kept in the customer's own infrastructure, and audit logs stands in the opposite position, and that is the advantage of doing it properly.
Sources:
- Baidu Baike, the entry on Forward Deployed Engineer (checked October 16, 2026; the site blocks automated access, so no link is embedded).
- Tencent Cloud ecosystem report "Industry Observation and Practice of the FDE Model" (PDF): kexinjinke.com/downloads/fde-industry-observation-and-practice.pdf
- "FDE technical framework for SMEs" v3, a three-tier delivery framework circulating in China's FDE community: scribd.com/document/1060966517
- AWS × Strand Partners, "Unlocking Vietnam's AI Potential" 2026 (VnEconomy, September 30, 2026): vneconomy.vn
- Deloitte × Meta, Vietnamese SMEs lead the region in AI adoption (BBW): bbw.vn
- Tuổi Trẻ (August 18, 2026), how Vietnamese SMEs use AI and how many reach maturity: tuoitre.vn
- VnEconomy citing the UEH survey in GIZ's 2025 annual report, cost barriers for SMEs: vneconomy.vn
- Reference price lists by tier from China's SME FDE workshops: datakrs.com and shixian.com/jobs/2057120159
- G-Company OS pricing (Vietnamese price benchmark): 5ac.vn/en/pricing
Want one person plus AI to carry what a whole delivery team used to? G-Company OS from 5ac ships in tiers with public pricing. See G-Company OS pricing.
— James Marcus, Content Lead 5ac.vn, October 2026. This is an analysis for business owners, not legal advice.