Start here: your first AI workflow for an ecommerce store

If you’ve poked at ChatGPT a few times but never actually shipped an AI workflow inside your ecommerce store, this is the place to start. One workflow. Two hours of setup. Real impact.

We’re going to build the simplest, highest-ROI starter: AI-drafted product description rewrites for your top 30 SKUs.

It works on Shopify, WooCommerce, BigCommerce or Magento. It needs zero engineering. It takes one afternoon. And it pays back immediately.

Why this workflow first

Three reasons:

1. High leverage. Product copy directly affects SEO and conversion. Both move within weeks.
2. Low risk. Nothing customer-facing changes until you approve it. No automation runs without your review.
3. Reusable foundation. The same prompt structure scales to your full catalog later.

What you need

• ChatGPT Plus or Claude Pro ($20/month). Either works for this.
• A spreadsheet (Google Sheets is fine).
• 90 minutes of focus.
• Admin access to your store.

That’s it. No n8n, no automation tools, no APIs. We’ll build it manually first; automation comes later.

Step 1: Pick the 30 products that matter most

Don’t spread thin. Pick the SKUs where copy improvements have the highest impact:

• Top 10 best-sellers (where conversion lifts compound).
• Top 10 SEO opportunities (where you rank page 2-3 and could rank top-5).
• 10 wildcards from underperforming product pages with traffic.

Total: 30 products. Don’t let yourself add more.

Step 2: Pull product data into a spreadsheet

For each product, fill in:

• SKU
• Product title
• Category
• Top 3 specs (material, dimensions, anything that matters)
• Target audience (one short phrase: “first-time skincare buyers, 25-35”)
• Focus keyword (the phrase you want to rank for)

If you don’t have a focus keyword yet, search the product type in Google. Look at the autocomplete suggestions and “people also ask” boxes. Pick something specific.

This step is the most important. Garbage in, garbage out.

Step 3: Write your brand voice prompt

In your AI tool of choice, create a Project (ChatGPT) or a Project (Claude). Inside the project’s custom instructions, paste a 5-10 line description of your brand voice.

Example for an apparel brand:

You write product descriptions for a slow-fashion apparel brand.

Voice:
- Sentences average 12 words. Never longer than 20.
- Always opens on a feeling or use case, never on fabric or specs.
- Uses contractions, occasional dry humor, second person rarely.
- Never uses: "premium", "elevated", "essential", "wardrobe staple", "high-quality".
- Never uses exclamation marks.
- Audience: style-led adults 28-45 who don't want to think too hard about getting dressed.

Add 2-3 examples of existing descriptions you wrote and are proud of. AI imitates examples better than instructions.

Step 4: The reusable product prompt

Inside that same project, paste this template:

Write a product description for an ecommerce store.

Product: {title}
Category: {category}
Top specs: {specs}
Audience: {audience}
Focus keyword: {keyword}

Requirements:
- 90-130 words
- First sentence: hook on a use case or feeling, not specs
- Middle: 3 benefit bullets
- Last sentence: a soft CTA, no exclamation marks
- Natural use of the focus keyword and one synonym
- Follow the brand voice in the project context

Now, for each of your 30 products, fill in the curly-brace fields and run it. AI gives you a draft.

Step 5: Review and edit

You’re not skipping this. The whole point is human + AI, not AI alone.

For each draft:

• Read it once.
• Fix any factual errors (wrong material, wrong dimensions).
• Tweak one or two sentences if the voice slips.
• If it’s mostly fine, ship it.

Budget 2-3 minutes per product. That’s 60-90 minutes for all 30. Add the prompt time, and you’re at 2 hours total for 30 production-ready descriptions.

Step 6: Push to your store

For Shopify: copy and paste, or do a CSV update via Matrixify. For WooCommerce: paste in the WP admin, or CSV via WP All Import. For BigCommerce: bulk edit via the admin or CSV import. For Magento: same patterns, slightly different tooling.

Don’t push all 30 at once. Push in batches of 10. Watch your analytics. Make sure nothing broke.

Step 7: Measure

Two weeks after going live, check:

• Search Console: impressions and clicks on the affected product URLs. You’re looking for upward movement.
• Time on page: should rise (Plausible, GA4, whatever you use).
• Conversion rate: usually a slow lift over 30-60 days for catalogs of this size.

Don’t expect overnight miracles. The signals stack over weeks.

What to ship next

Once this works, the natural next steps:

• Scale to your full catalog using the same prompt structure but with n8n or Make for batch runs.
• Add metadata (titles and meta descriptions) using a similar prompt.
• Add FAQ blocks to category pages using customer review data.
• Tackle helpdesk with a separate prompt and your support tool’s API.

Each of these is a separate workflow. None of them are necessary until the first one is working.

Common questions

“What if I don’t like the AI output?” Tune the prompt. Add an “avoid” list. Add more examples in the brand voice block. Iterate, don’t escalate to “AI doesn’t work.”

“What if I have 5,000 products?” Start with 30. Prove the workflow. Then scale via Make, n8n or Zapier with the same prompt.

“Will Google penalize this?” Not if you review the output. Google penalizes thin, useless content, regardless of origin. AI-assisted content with human review and accurate facts ranks fine.

“How long until I should expect results?” Time-on-page moves in 1-2 weeks. SEO impressions in 4-8 weeks. Conversion rate in 4-12 weeks for catalogs this size.

If you’re stuck

The hard part isn’t the AI. It’s deciding what to ship first and committing 2 hours to ship it. If you’d rather have someone walk through your store and pick the workflow with you, the free AI audit does exactly that.