E-commerce product description and image enhancement workflow

Introduction

Product descriptions and images are the backbone of e-commerce conversion rates. A customer scrolling through your shop will either pause to read your description and examine your images, or they'll move on to a competitor. Yet most sellers treat these assets as a one-time setup task rather than an ongoing optimisation opportunity.

The problem is clear: manual enhancement is exhausting. You've got hundreds or thousands of products. Each one needs a better description written (or rewritten), images tagged with alt text, and metadata polished. You could hire someone to do this, but then you're hiring for a task that happens once per product per quarter. The maths doesn't work.

This is where combining AI-Boost and SellerPic-AI solves the problem elegantly. AI-Boost handles text generation and refinement; SellerPic-AI handles image analysis and enhancement. Together, they can process your entire catalogue automatically. With the right orchestration layer, you set this up once and forget about it. New products get enhanced on arrival. Existing products get refreshed on a schedule. No manual handoff.

The Automated Workflow

We'll use Make (formerly Integromat) to orchestrate this workflow. Make is ideal here because it handles REST API calls smoothly, has good rate limiting controls, and integrates well with both tools without requiring custom code. If you prefer open-source infrastructure, n8n works identically; we'll note the differences as we go.

Step 1:

Trigger on New or Updated Products

The workflow starts with a trigger. You have three options:

1. Trigger on a schedule (check for new products every 4 hours)

2. Trigger via webhook (your e-commerce platform sends a notification when products are added)

3. Trigger via a polling module (Make queries your product database periodically)

For this example, we'll assume you're pulling products from your e-commerce platform via API. Most platforms (Shopify, WooCommerce, custom stores) expose a products endpoint. You'll need your API credentials.

In Make, create a new scenario and select the HTTP module. Configure it to list products:

Copy
GET https://your-store-api.com/v1/products?status=active&limit=50&offset=0
Headers:
  Authorization: Bearer YOUR_API_TOKEN
  Content-Type: application/json

Make will return a JSON array of products. Each product object should contain:

• product_id
• product_name
• current_description
• image_url (primary image)
• category

If your API returns pagination, use Make's iterator module to loop through all results without hitting limits.

Step 2:

Generate Enhanced Description with AI-Boost

Once you have product data, pass it to AI-Boost. AI-Boost's API takes your current description and category, then returns an enhanced version optimised for search and conversion.

Configure a new HTTP module in Make:

Copy
POST https://api.ai-boost.io/v1/enhance-description
Headers:
  Authorization: Bearer YOUR_AIBOOST_API_KEY
  Content-Type: application/json

Body:
{
  "current_description": "{{product.current_description}}",
  "product_name": "{{product.product_name}}",
  "category": "{{product.category}}",
  "tone": "professional",
  "target_length": 150,
  "include_keywords": true
}

AI-Boost responds with:

Copy{
  "enhanced_description": "Premium wool blend winter coat...",
  "keywords": ["winter coat", "wool blend", "thermal insulation"],
  "readability_score": 8.2,
  "estimated_conversion_lift": "12-15%"
}

Store the enhanced_description in a variable. You'll write this back to your store later.

Step 3:

Analyse and Enhance Image with SellerPic-AI

In parallel (or sequential, depending on your preference), send the product image to SellerPic-AI. This tool extracts metadata, suggests alt text, detects image quality issues, and can even generate complementary crop suggestions.

Create another HTTP module:

Copy
POST https://api.sellerpic-ai.com/v1/analyse-image
Headers:
  Authorization: Bearer YOUR_SELLERPIC_API_KEY
  Content-Type: application/json

Body:
{
  "image_url": "{{product.image_url}}",
  "product_id": "{{product.product_id}}",
  "generate_alt_text": true,
  "analyse_quality": true,
  "detect_objects": true
}

SellerPic-AI responds with:

Copy{
  "product_id": "12345",
  "alt_text": "Navy wool blend winter coat with thermal lining, size medium",
  "quality_score": 8.7,
  "quality_issues": [],
  "detected_objects": ["coat", "model", "fabric"],
  "suggested_crops": [
    {
      "x": 50,
      "y": 100,
      "width": 400,
      "height": 600,
      "use_case": "thumbnail"
    }
  ],
  "metadata": {
    "colour_palette": ["navy", "grey", "white"],
    "lighting_quality": "excellent",
    "background_clarity": "clean"
  }
}

Store the alt_text and quality metrics.

Step 4:

Conditional Logic: Update Only If Improved

This is important. Not every product needs updating. If AI-Boost's enhancement is marginal or if the image quality is already excellent, you should skip the update to save API costs and avoid unnecessary churn.

In Make, add a router module with conditional logic:

Copy
Condition 1: AI-Boost readability_score > 7.0 AND estimated_conversion_lift > 5%
  Action: Proceed to update description

Condition 2: SellerPic-AI quality_score < 7.5
  Action: Flag for manual review (we'll handle this in step 6)

Condition 3: Both conditions met
  Action: Proceed to step 5

If conditions aren't met, the workflow logs the product and stops, avoiding wasted writes.

Step 5:

Write Enhanced Data Back to Store

Once you've validated the improvements, write the enhanced description and alt text back to your e-commerce platform.

Configure an HTTP module for your store's update endpoint:

Copy
PUT https://your-store-api.com/v1/products/{{product.product_id}}
Headers:
  Authorization: Bearer YOUR_STORE_API_TOKEN
  Content-Type: application/json

Body:
{
  "description": "{{enhanced_description}}",
  "image_alt_text": "{{alt_text}}",
  "meta_keywords": "{{keywords}}",
  "updated_by": "ai-automation",
  "updated_at": "{{now}}"
}

Your store API should return a 200 status code. If it returns 400 or 500, Make's error handling will catch it and retry.

Step 6:

Logging and Exceptions

Create a Google Sheets or Airtable module that logs every run. This serves two purposes: audit trail and exception tracking.

For each product processed, log:

• Product ID
• Original description length
• Enhanced description length
• AI-Boost readability score
• SellerPic-AI quality score
• Timestamp
• Status (updated, skipped, error)

If an API fails or a quality check fails, send a Slack notification:

Copy
POST https://hooks.slack.com/services/YOUR/WEBHOOK/URL

Body:
{
  "text": "Product {{product.product_id}} failed quality check",
  "blocks": [
    {
      "type": "section",
      "text": {
        "type": "mrkdwn",
        "text": "*AI Enhancement Alert*\nProduct: {{product.product_name}}\nReason: {{error_message}}\nAction: Review manually"
      }
    }
  ]
}

Scheduling and Frequency

Set the entire scenario to run on a schedule. A sensible approach:

• Run a "new products" check every 6 hours (catches recent uploads)

• Run a "refresh all products" check weekly or monthly (improves descriptions over time as AI models improve)

For large catalogues (10,000+ products), split the work:

• Monday: Products A-D

• Tuesday: Products E-H

• Wednesday: Products I-L

• And so on

This distributes API calls and avoids hitting rate limits.

The Manual Alternative

If you prefer more control over which products get enhanced, or you want to review AI output before publishing, use a simpler workflow:

1. Run AI-Boost and SellerPic-AI on a batch of products (e.g., 100 at a time)

2. Store the results in a staging table (Google Sheets, Airtable, or a database)

3. Review the output manually or with a colleague

4. Approve or reject each enhancement

5. Only then write approved changes to your store

This takes longer but gives you full visibility. It's useful for high-value products or when you're first testing the workflow. Once you're confident in the output quality, switch to full automation.

Pro Tips

Rate Limiting and Cost Control

Both AI-Boost and SellerPic-AI have rate limits. Check their documentation. Most offer 100-500 requests per minute on standard plans. If you have 5,000 products, don't try to process them all in one go. Batch them: 50 per hour, or 100 per day. Use Make's delay module to space out requests.

Copy
Add a delay of 5 seconds between each product:
Delay module: 5 seconds

This sounds slow but prevents throttling and keeps your costs predictable.

Error Handling and Retries

Make's error handling is decent out of the box, but you should be explicit. Configure a "catch" clause on HTTP modules to handle 429 (rate limit) and 500 (server error) responses:

Copy
If status = 429: Wait 60 seconds, retry once
If status = 500: Log error, skip product, continue
If status = 400: Log error and notify via Slack (likely a data issue)

This prevents your entire workflow from failing because one API is temporarily unavailable.

Quality Thresholds

Set sensible thresholds before updating. For descriptions, only accept enhancements with a readability score above 7.0. For images, only accept alt text if confidence is above 90%. These thresholds prevent poor-quality updates from going live.

Testing with a Subset

Before running this on 10,000 products, test it on 10. Create a test scenario and run it manually. Review the output. Tweak your prompts or thresholds. Once you're happy, scale up.

Monitoring Costs

Track your API usage. AI-Boost typically charges per description (£0.01–0.05 depending on plan). SellerPic-AI charges per image analysis (£0.02–0.10). For 1,000 products, that's £30–150 for a full refresh. Budget accordingly. Most tools offer discounts on larger volumes, so ask.

Cost Breakdown

If you process 100 products per month, costs are around £30–40. If you process 5,000 products per month, invest in higher-tier plans and expect £150–250.

Putting It All Together

The workflow runs like this:

1. Trigger fires (schedule or webhook)

2. Fetch batch of products from your store

3. For each product, call AI-Boost (enhanced description) and SellerPic-AI (image analysis) in parallel

4. Check quality thresholds

5. Write updates back to store

6. Log results and send alerts if errors occur

7. Schedule the next run

No manual work. No copy-pasting. No forgotten products. Every product in your catalogue benefits from consistent, intelligent enhancement.

The first run takes time to set up. But after that, it scales instantly. Add a new product, and it gets enhanced automatically. Update a product, and the workflow refreshes its description and metadata. Your e-commerce storefront improves continuously, without you lifting a finger.