Can AI Agents Run a ₹11,000 Shopify App in 90 Minutes a Week?
FeedMedic is a low-cost, mostly agent-run idea: a Shopify app that fixes Google Merchant Center disapprovals. Here's how it works, what it costs in rupees, and why our critic scored it only 6.5/10.
हिंदी में पढ़ेंThe email every store owner hates
Picture a small online store owner opening an email from Google that says “Products disapproved.”
There’s no phone number to call and no clear explanation. Just a list of error codes like “missing identifiers” and “mismatched value”, and meanwhile their products have quietly dropped out of Google Shopping.
That small, dull, recurring problem is the basis for a business idea we’ve been testing: FeedMedic, a Shopify app built so that AI agents do almost all of the work.
By our estimates, it costs about ₹11,200 to start, around ₹10,300 a month to run, and needs roughly 1.5 hours a week of your time at the beginning.
It sounds very appealing. It’s also not a sure thing, and we’ll explain why further down.
The problem: invisible products mean lost sales
Shopify merchants who sell through Google Shopping ads or Google’s free product listings send their product data to Google Merchant Center. When Google doesn’t like something in that data, it disapproves or “limits” the product.
When that happens the merchant loses traffic, and often doesn’t realise it.
The reasons are often cryptic:
- Missing product identifiers (GTIN/MPN)
- Price or availability that doesn’t match the website
- Promotional text in titles (“BEST DEAL!!!”)
- Image problems
- Missing shipping or return information
Fixing these usually means editing hundreds of products by hand or paying an agency. Shopify’s own Google & YouTube channel shows the errors, but it does little to fix them.
The target customer is a founder-run Shopify store with 50 to 10,000 products and no feed specialist on the team.
How the business works
FeedMedic is a Shopify app. A merchant installs it, connects their Google account, and the app’s agent:
- Finds out why products are being rejected
- Writes compliant fixes using only the store’s own product data
- Pushes those fixes back once the merchant approves them
Billing runs through Shopify’s Billing API. Charges appear on the merchant’s Shopify invoice and Shopify pays out to your bank, so you don’t need a separate payment gateway.
Who it’s for: this is a global product. Your customers are Shopify merchants anywhere in the world, not just in India.
How the AI agents run it, step by step
Here’s the daily workflow, almost all of it handled by agents:
- Install: The merchant installs from the Shopify App Store and connects Google through OAuth.
- Scan: The agent pulls product statuses and issues from Google’s Merchant API and matches them to the Shopify product data.
- Diagnose: It groups issues by root cause, ranks them by impact and explains each in plain English. It also says clearly what it can’t fix, such as account suspensions, and offers a step-by-step guide to Google’s appeal process instead.
- Generate fixes: Claude (Haiku for classification, Sonnet for rewrites) proposes changes such as cleaning up titles, mapping barcodes that already exist in the store to GTIN fields, and filling in category, colour, size and gender from variant data. It never invents specs or claims.
- Approve: The merchant sees a before/after table and approves in bulk. Growth-plan users can auto-approve low-risk fixes.
- Apply: Changes go to Shopify or to a separate “supplemental” Google data source, so the storefront text stays untouched.
- Monitor: Each day the agent re-checks statuses, catches new problems and sends a weekly email summary.
- Support: An in-app chat agent answers questions using the docs and the merchant’s own diagnostic data. Difficult cases go to you with a pre-written summary.
- Maintain: The agent watches Google and Shopify API changes, opens code tickets and writes your weekly metrics report.
For marketing, the agent also maintains a free public “Merchant Center error library”: one plain-English page per error code, each with a “fix this automatically” link to the app.
The money (all estimates)
Start-up cost: about ₹11,200 This includes the ₹1,800 Shopify Partner registration fee plus domain and setup. Hosting (Cloudflare Workers or Fly.io), the database (Neon Postgres), email (Resend) and error tracking (Sentry) all start on free or low tiers.
Monthly running cost: about ₹10,300 This is mostly Claude API usage plus hosting. Your first month comes to roughly ₹21,500 in total.
Pricing:
| Plan | Price | What’s included |
|---|---|---|
| Free | ₹0 | Full scan + 10 fixes |
| Starter | ~₹1,800/month | Up to 1,000 products, unlimited fixes, daily monitoring |
| Growth | ~₹4,600/month | Up to 10,000 products, auto-apply rules, supplemental feed |
Paid plans come with a 7-day free trial.
Break-even: On these estimates, about 6 Starter customers or 3 Growth customers would cover monthly costs.
Proof that people pay: Established Shopify feed apps such as Simprosys, AdNabu and Multiify have thousands of reviews and charge roughly ₹470 to ₹7,500+ a month.
Note for Indian founders: You’d be earning from overseas merchants, so speak to a CA about GST on export of services, how to receive Shopify payouts, and whether to set up as an individual or a company. AI API costs are billed in dollars, so rupee swings will affect your margins.
What the human still does
“Almost entirely AI-run” doesn’t mean zero work. You still have to:
- Build it once: about 20–30 hours with an AI coding agent, then testing on a development store.
- Get approved: Shopify App Review usually takes 1–3 weeks and sometimes 2 rounds. Google’s OAuth verification needs a privacy policy, a homepage and a demo video (the agent can script the video).
- Own the accounts: Shopify Partner, Google Cloud, domain and bank.
- Weekly (~1 hour): read the metrics report, spot-check 15 random AI rewrites for accuracy, and handle escalated tickets.
- Every few months: approve and deploy updates when Shopify or Google change their APIs.
Risks and honest verdict: 6.5/10
Our critic gave FeedMedic 6.5 out of 10 with a verdict of “maybe.” The scorecard:
- Capital: 8/10
- Demand: 7/10
- Scalability: 7/10
- Legal risk: 7/10 (higher is safer)
- Automation: 6/10
- Competition: 3/10
Competition is the big problem. Simprosys, AdNabu, Multiify, Mulwi, DataFeedWatch, Channable and others are well established with large review counts. A new app with zero reviews ranks poorly. Google itself keeps adding AI tools such as Product Studio inside Merchant Center.
The timing advantage has mostly gone. Google’s move from the Content API to the new Merchant API in 2026 looked like a chance to win customers from tools that fell behind. By now, most established apps have migrated.
Fewer problems are fixable than it seems. The most common error, missing identifiers, often can’t be fixed honestly. The app can’t make up barcodes, and marking branded goods as having “no identifier” risks misrepresentation. Many “mismatched value” errors are technical problems, not copywriting problems.
There’s more human time than the headline suggests. The 1.5 hours a week only holds at very low install counts. Expect 3–5+ hours a week once you pass 100 merchants, because suspension tickets are emotional and need a person.
You depend on two platforms. Shopify can delist or copy you, and Google can change its API or build the feature itself.
Revenue will take longer. The plan says 6 weeks to first revenue. Once you allow for review and verification queues, 10–14 weeks is more realistic.
Bottom line: the demand is real and the costs are tiny, but you’d be a newcomer in a crowded market. It’s only worth pursuing if you find a narrow angle the established apps miss, such as policy fixes to titles and attributes for “limited performance” products.
The first three steps to start
1. Validate before you code. Have an AI agent analyse the 1–3★ reviews of the top 10 Google feed apps and the most common Merchant Center Help Community threads. Sort the issues into three groups: fixable from existing store data, needs new facts from the merchant, and technical or account-level. If fewer than about 40% are fixable from store data, or the established apps already handle them, stop.
2. Launch the error library first. Put up the free plain-English error-code site with a waitlist. It tests search demand and builds a launch list for roughly ₹0.
3. Start the slow paperwork early. Register your Shopify Partner account and begin Google OAuth verification now. Build a narrow first version that writes only to a supplemental Google data source and never to live storefront data. That means less breakage and less support work.
The idea isn’t a goldmine, but it’s a cheap, low-risk experiment if you go in with realistic expectations.
Figures converted at 1 USD = ₹93.37 (Fri, 20 Nov 2026 00:02:32 +0000). Costs are estimates.
These are business ideas, not financial, legal or tax advice. Check the facts, costs and rules yourself before you spend money.