amberstack/

case study

Daily social drafts for a one-person Shopify store

In-house pilot, Lithuanian e-commerce brand

stack
n8n, Claude, Shopify Admin API
build time
~25 engineer-hours
result
Posts/week: ~0 → 2-3. Time per post: hours of staring at the blank page → up to 15 min of proof, translation, and image selection.

This is an in-house pilot. The store is family-run, so I had a first user who actually wanted it and could ship without approval rounds. Treat it as an in-house example, not a paid client win. Names and product details are kept private on purpose.

The problem

The store is one person running everything: stock, photos, packing, customer service, the lot. Social was supposed to drive traffic. Social was not happening.

The slow part wasn't typing. It was deciding what to say. "What do I post today" took longer than the post itself, so most days the answer was "nothing." The brand has plenty to draw on, Lithuanian name days, holidays, seasons, weather, and a catalog of pieces with a real story behind each. But putting those together from scratch every morning is hard to do alone.

What I built

A daily n8n workflow:

  1. Pulls the day's context from a public Lithuanian calendar source (date, name day, holiday, season, weather, brief cultural notes).
  2. Fetches all products from the Shopify Admin API. Filters out items with zero stock and items that aren't published.
  3. Sends both to Claude with the brand's voice rules and a few post formats to follow.
  4. Outputs 2-3 drafts per run. One tied to the calendar. One or two tied to a specific in-stock piece that fits the day.
  5. Each product draft gets a tracking link built at draft time, so the owner can see later which drafts actually drove clicks.
  6. Drafts arrive in the owner's morning inbox as a clean list. Pick one, edit, ship.

The output shape, roughly:

{
  "draft_number": 2,
  "hook_style": "product",
  "text": "...",
  "product_id": "gid://shopify/Product/[redacted]",
  "product_url": "https://[redacted]/products/[redacted]?utm_source=facebook&utm_medium=organic&utm_campaign=2026-05-24",
  "sources_used": "holiday, season, weather_today, product"
}

What a calendar-hook draft looks like: a short scene tied to the day's holiday or name day. What people are doing outside, what the season feels like, where the brand's pieces quietly fit in. No product, no link. Pure mood.

What a product-hook draft looks like: the same scene, but built around one specific piece in stock that suits the day. One line on what the piece is, a small detail that ties it to the day, a tracked link.

Results

  • Posts per week: roughly 0 → 2-3.
  • Time per published post: hours of staring at the blank page → up to 15 min, mostly translation polish, image selection, and a final read.
  • Build: 10h to working prototype, ~25h total to ship-ready (polish, error handling, tone tweaks, link tracking).
  • Side effect: even on skipped days, the drafts pile up as a ready-made idea bank. The owner reuses older drafts when a similar kind of day rolls around again.

The interesting number isn't "saved X hours per week." She wasn't spending hours per week on this before, she was spending zero. The win is moving the activity from "not happening" to "happening regularly," which is a different kind of result than saving time on work that was already getting done.

What it cost to run

One run per day. Using top-tier models (no swapping in a cheaper one, because draft quality is the whole point), the LLM bill comes in at ~€0.17/run, ~€5/month at one run per day.

The calendar feed is free. The Shopify Admin API is free for store owners. n8n runs on a shared Hetzner box that's already paid for by other internal work, so the only extra cost for this workflow is the LLM bill, nothing more.

Roughly under €10/month all in, and every piece of it is something the owner can swap out without rebuilding the whole thing.

Back to work