How Ordinary People Can Use AI to Break Into Short Video and Make Money

One account, 20 videos, all produced by AI. The worst got hundreds of thousands of views. The best: 3 million.
20 videos, every single one exceeding 100k views.
Using AI to cut video production costs from 500 yuan to 5 yuan.
I’ve written about an AI short video workflow before — two accounts with 300k followers each, generating 150 million GMV.
But today, I don’t want to talk about his account data.
I want to write about something I witnessed firsthand sitting in his office.
I asked him: Can an ordinary person replicate this workflow?
He said: Yes.
This article will lay out that answer clearly.

First, What’s Happening in the Industry

Just this week, Snap announced 1,000 layoffs and closed over 300 job positions.
The reason came down to one sentence: AI’s rapid development allows smaller teams to achieve the same output.
Over 65% of Snap’s new code is now generated by AI, saving an estimated $500 million annually.
Lay off 1,000 people, save $500 million.
This isn’t unique to internet companies — it’s happening across the entire content industry.
The short video space is no exception.

Previously, the production cost for one sales video was: 500 yuan for script planning, 300 yuan for shooting and editing, and 200 yuan for operation follow-up. A video could take two days at best, a week at worst.
Now?
AI writes the script in 10 minutes. AI-assisted editing takes 30 minutes. One person can produce a finished video from scratch in half a day.
Costs drop from 1,000 yuan to under 100 yuan.
This isn’t theory — it’s the reality running at Mr. Wang’s company.

═══════════════════════════════

「Snap laid off 1,000 people using AI and saved $500 million — where did that money go?」

═══════════════════════════════

What I Saw at Mr. Wang’s Company

Mr. Wang’s company is not big.
Wuhan, an ordinary office, about a dozen people.
But what they’re doing is more efficient than many teams of a hundred.

One Douyin account focused on private domain sales.
Over the past three months, AI wrote scripts for 20 videos.
I asked him: Out of these 20, what was the worst performance?
He said: Over 100k views.
And the best?
2 million. A few broke 3 million.
I was quiet for a moment.
Not because the numbers are big.
But because those 20 videos — from script to editing — were almost entirely run by AI.
The human role was judgment: choosing the topic, reading the data, deciding where to go next.
Execution was left to AI.
That was my deepest takeaway after sitting in his office for a day.

How This Workflow Actually Runs

Divided into four parts.

Part 1: AI Writes Scripts

Private domain sales videos follow a fixed structure:
Opening hook (capture attention in first 3 seconds) → Pain point amplification (make the user think “that’s me”) → Product placement (natural, not forced) → Call to action (guide to DM or click).
This structure can be mass-produced by AI.

Mr. Wang’s approach: break down the structure of viral videos, feed it to AI along with the product’s selling points and target audience, and have AI write scripts following that structure.
Generate 10 candidate scripts at once, then people pick 3 to shoot.
Previously, a copywriter could produce 2-3 scripts per day, not all usable. AI generates 10 at once, with consistent quality and 10x faster speed.

Part 2: AI-Assisted Editing

What AI can do now for editing:

• Automatically identify key sentences in the video and mark time points.
• Automatically match background music rhythm with scene transitions.
• Automatically generate subtitles with over 95% accuracy.
• Automatically crop and resize for different platforms.
  What humans need to do is final tweaks and reviews.
  A 60-second sales video used to take 2-3 hours to edit. Now, editors using AI assistance produce a finished version in 45 minutes.
  3x faster, with no drop in quality.

Part 3: Driving Traffic to Private Domain

After the video is posted, traffic flows into the private domain.
In this step, AI handles:

• Automatically monitoring comments to identify purchase-intent messages.
• Generating personalized reply responses — not copy-paste, but tailored to each comment.
• Sending DMs at key moments to guide commenters into WeChat.
  One operator could previously manage at most 3 accounts. Now, with AI, one person can run 5-8 accounts simultaneously.
  Efficiency more than doubles.

Part 4: Private Domain Customer Maintenance

Once customers are in WeChat, how do you nurture them?
AI helps by:

• Automatically tagging and segmenting customers based on purchase history and interaction frequency.
• Pushing different content at different frequencies to different segments.
• Triggering actions at key moments: birthdays, 7 days after last purchase, alerting high-intent customers about new products immediately.
  With this system, repeat purchase rates can increase by over 30%.
  It’s not about piling on human labor — it’s about AI recognizing each customer’s state and saying the right thing at the right time.

═══════════════════════════════

「AI touches all four parts — that’s Mr. Wang’s company workflow」

═══════════════════════════════

20 Videos, Each Over 100k: What’s the Underlying Logic?

I asked Mr. Wang this question.
His answer was straightforward: AI handles execution, humans handle judgment.
Let me break it down.

First Layer: Topic Selection Determines the Ceiling

Twenty videos each exceeding 100k isn’t because AI writes well.
It’s because the topic direction was correct.
For private domain sales videos, topics have three dimensions:

• Is the pain point real enough?
• Is the product solution clear enough?
• Is the target audience precise enough?
  Judgment on these three dimensions comes from humans. What AI does is efficiently execute once the direction is set.
  If the topic is wrong, AI writing 100 scripts won’t help. If the topic is right, AI generates 10 options, and you only need to shoot the best 3.

Second Layer: Data Is the Commander

After the 20 videos go live, each one’s data differs.
Completion rate, like rate, comment rate, DM conversion rate.
Mr. Wang’s team checks these four metrics daily.
Which video has a high completion rate? The opening hook worked — reuse that structure next time.
Which video has a high DM conversion rate? The call to action was effective — extract that wording for reuse.
Which video performs poorly? Find the reason, change one variable, test again.
AI provides enough ammo to quickly test and iterate.
Previously, with 2 videos a week, the testing cycle was too long — by the time you found the direction, the traffic window might have closed.
Now, with 10 videos a week, post 3, look at the data, optimize the next batch.

Third Layer: Private Domain Is the True Moat

Douyin traffic belongs to the platform.
Private domain customers belong to you.
The core value of this workflow isn’t making videos go viral — it’s turning every person attracted by a video into an asset you can repeatedly reach.
A video with 3 million views that only brings platform metrics, with no private domain accumulation, means the next video starts from zero.
But if every video drives traffic into private domain, 100k views bring in 5,000 people. Those 5,000 people remain in WeChat, can be sold to a second, third, fourth time.
Traffic is fuel; private domain is the engine.

How Ordinary People Can Enter With This Logic

Mr. Wang’s workflow requires team coordination and deep understanding of private domain sales — not everyone can copy it directly.
But the underlying logic can be used by anyone.

Path 1: Create a Vertical Account, Mass-Test Topics With AI

Pick a field you’re familiar with.
It doesn’t have to be AI. It could be an industry you’ve worked in, a place where you’ve stepped on landmines, anything you know more about than the average person.
Use AI to mass-produce content in that field, post 5-7 videos per week, track data, find the viral formula, then replicate.
The key isn’t how well AI writes — it’s how good your judgment is in that field.
AI is an amplifier; your judgment is what gets amplified.

Path 2: Help Offline Business Owners Set Up This Workflow, Charge Service Fees

Most offline business owners don’t know how to set up Mr. Wang’s workflow.
But they have products, they have customers — just missing the system.
You help them build: AI script templates, editing workflow, private domain traffic scripts, customer maintenance SOPs.
Setup fee: 5,000-20,000 yuan. Monthly maintenance fee: 2,000-5,000 yuan.
One client brings 20,000-80,000 yuan per year. Sign 3 clients, and that’s your baseline income.
You don’t need to be Mr. Wang — you just need to know more about AI than your clients.

Path 3: Partner With Those Who Have Content Foundation — You Provide AI System, They Provide Content

This is what Mr. Wang and I are doing now.
He has short video content experience and private domain sales know-how.
I have AI toolchains and a content factory logic.
The two of us collaborate: set up a new account, combine his industry experience with my AI system, run a new monetization path.
1+1 doesn’t equal 2 — it equals 10.
This type of partnership will become more common in the AI era. If you have depth in a certain field, find someone with AI skills, or vice versa.
Each brings half, filling what the other lacks.

═══════════════════════════════

「One account, 20 AI videos, max 3 million views — these are real numbers」

═══════════════════════════════

What I’m Doing in Wuhan

I came back from Pattaya not for tourism.
I came to see, after this workflow is proven, what the next step can be.
Mr. Wang has validated the AI production workflow for short videos.
I’ve validated the content factory logic.
The two systems overlap and complement each other.

What we’re doing now is integrating these two systems into one, specifically serving offline entrepreneurs who want to use short videos for private domain sales but don’t know how to leverage AI for efficiency.
This isn’t a training course.
It’s a hands-on service: we help set up the workflow for your account, get it running, track data, optimize, until it works.
The value isn’t the tools — it’s the methodology that has already been proven.

How long did it take Mr. Wang to validate this workflow?
Nearly a year.
How many pitfalls, how much tuition did he pay to get 20 videos each exceeding 100k?
Only he knows.

That’s why “finding a professional to do it” is a form of judgment, not laziness.

Three Rules

Rule 1: AI is execution layer; your judgment is decision layer. Let AI do what it should do, and leave your time for judgments only you can make.

Rule 2: Data is the only thing you should trust. A video that feels good may not have good data; a video with good data is the truly effective direction.

Rule 3: Private domain is the end goal; traffic is the means. Every video should answer: what’s the purpose of this video? Does it guide people into private domain? After they enter, how do you nurture them?

Summary in Three Lines

1. AI penetration in the short video industry has already happened — not a trend, but reality. Mr. Wang’s account produced 20 AI videos, every single one over 100k views — that’s verifiable data.

2. The core of this workflow isn’t the tools — it’s topic judgment + data review + private domain accumulation. Tools are available to everyone; your judgment is your moat.

3. Three paths for ordinary people: build your own vertical account, help offline business owners set up the workflow, or partner with someone who has industry depth. You don’t need to start from scratch — find the right direction and move fast.

═══════════════════════════════

═══════════════════════════════
Sitting in Mr. Wang’s office that afternoon, looking at the data, thinking about how to replicate this workflow for more business owners.