While everyone on social media was trying to become a "prompt engineer" or launch a viral course on "AI for Creatives," a quiet, unglamorous job just kept printing money: writing product descriptions.
It is the least sexy, most reliable AI-leveraged business you can run.
Why? Because it's not a creative job. It's a logistics job. It's a conversion job. It's a boring, expensive problem that every single e-commerce store has.
And there are millions of them.
The Problem Hasn't Changed
Shopify, Amazon, and Etsy are still drowning in terrible copy. Go to any mid-level e-commerce site right now. You'll find one of two things:
- The original, cryptic, often badly translated manufacturer description. (e.g., "Premium fabric material good for use.")
- The obvious, first-draft AI slop from when the owner tried ChatGPT in March and got this: "Unleash your potential with our meticulously crafted T-shirt..."
Both are sales-killers. Business owners know this. They know that a good product description can be the difference between a 1% and a 4% conversion rate. They know it's the difference between a click and a sale.
They just don't have the time or the will to fix 250 of them.
This isn't a problem of passion. It's a problem of volume. They don't need a poet. They need a plumber. They need someone to come in, fix the leaks, and get the system working.
You Are Not a Writer. You Are an Assembly Line.
This is where you come in. Your job isn't to be a "writer." Your job is to run a high-efficiency assembly line.
Here's the 2025 workflow. It is not glamorous.
- Intake: Client sends you a Google Sheet with 100 products, their features, and their target audience.
- Prompting (The 10%): You feed this data into your AI of choice using a refined prompt template you've spent months perfecting.
- Curation (The 90%): You get 100 AI-generated drafts. 80 of them are decent. 20 are garbage. You review every single one.
- Editing: This is the money. You're not just correcting grammar. You're injecting the brand's voice. You're turning bland "features" (e.g., "1.5-inch heel") into tangible "benefits" (e.g., "The perfect 1.5-inch heel for all-day comfort without sacrificing style.").
- Delivery: You send back the completed Google Sheet, often in 48 hours.
The owner tried to do this. They got overwhelmed at product #4. The "prompter" you're competing against just sent back the raw AI slop, and the owner is furious.
You are the only one who delivered 100 sellable descriptions, on time, in the right voice, without making them do any work.
The Math on Boring
Stop charging by the hour. Hourly rates punish efficiency. The client doesn't care if it took you 10 minutes or 10 hours; they care about the 100 finished descriptions.
You charge for the solution.
And you charge in packages.
- 10-Description Test: $150 ($15/description)
- 50-Description Package: $600 ($12/description)
- 100-Description Package: $1,000 ($10/description)
Let's say you land one client who needs 100 descriptions. With your AI-leveraged workflow, you can finish that entire package in 6-8 hours of focused work. That's $125 - $160 per hour.
Now, do that for three clients a month.
That's $3,000. For a "boring" job that everyone else ignored because they were too busy trying to build a personal brand.
This isn't a hustle you have to post about. It's a quiet, profitable service business. You're not selling creativity. You're selling completion.
And in an economy of half-finished projects and terrible AI drafts, "done" is the most valuable commodity there is.
August 2026 update
Editor's note. Model names in the workflow above are from October 2025 and are now several generations out of date. The workflow itself doesn't depend on which model you use — pick the current frontier one. Everything else is unchanged.
The pricing in this essay has aged reasonably. What has moved is the client's baseline expectation: in 2025 you were competing against an owner who had tried ChatGPT once and given up at product #4. In 2026 you are more often competing against an owner whose storefront platform ships a description generator for free.
That kills the low end. The $10-a-description volume package is much harder to sell when the platform will produce something passable at zero marginal cost. What survives is the same thing that survived the content correction: someone who takes responsibility for whether the copy actually converts, in the brand's voice, at volume, on a deadline. The plumber framing still holds. The price floor moved up, not down — the cheap tier is gone, not the work.