Feed AI - Getting Started

Feed AI is a powerful tool within SmartFeeds that allows you to use artificial intelligence to make many different kinds of changes to your product data. By creating your own "Tasks" and "Cases," you can give the AI specific instructions to create new attributes, rewrite titles and descriptions, or even check the quality of your data.

This guide will show you the process of setting up your first AI Task using a practical example, from selecting your input data to launching the optimization and interpreting the results.

Prerequisites

Before you begin, you should have an active product feed imported into SmartFeeds. You should also be familiar with the general interface of the Product Flow. This is the central area where you manage your data. Understanding features like joining (to combine different data sources) and mapping (to standardize attributes) is key to getting the most out of SmartFeeds.

Looking for inspiration? Feed AI now includes a built-in Prompt Library directly in the interface when creating a case (ready-to-use prompts and schemas for both Text and Image tasks).

Table of Contents

1. Understanding the Feed AI Interface: Tasks and Cases
2. Understanding Feed AI Credits and Cost Estimation
3. Configuring Your First AI Task: A Step-by-Step Example
   • Part 1: Configure and Preview Your Task (Free for All Licenses)
   • Part 2: Activate and Use Your Results (Full Feature Access)
4. Quick Answers to Common Issues
5. Related Articles

Understanding the Feed AI Interface: Tasks and Cases

The core of Feed AI revolves around two main concepts. Think of it like organizing a project:

• A Task is like the main project folder. It holds everything for a specific optimization goal, including the input data and general settings (like the AI model). You can create different Tasks for different projects
  Example: one Task (or “folder”) for attribute extraction, another for title optimization
• A Case is like a specific instruction sheet inside your project folder. Each Task can contain one or more Cases. A Case is made of three parts that work together: a product filter, a prompt, and a schema. This allows you to apply different instructions to different segments of your products, all within the same project. Cases are processed in order; a product handled by the first instruction sheet will be excluded from the next.
  Example : within a single "Title Optimization" Task, you could have one Case with a prompt to create long, detailed titles for your "Electronics" category, and a second Case with a different prompt to create shorter, punchier titles for your "Sport" category.

From the main interface, you can also manage your Tasks to keep your workspace organized:

• Duplicate/Delete: You can duplicate or delete an entire Task by clicking the three-dot menu next to the Task name.
• Collapse Cases: To keep your workspace tidy when working with multiple Tasks, you can collapse the Cases of any inactive Task.

Understanding Feed AI Credits and Cost Estimation

Feed AI operates on a credit system to offer you maximum flexibility. Before diving into your first task setup, it's essential to understand how these credits are managed.

There are two types of credits:

1. Monthly Credits (Subscription)

These are renewed automatically each month via your subscription.

• Warning: They expire on the last day of the month. Be sure to use them!
• Recommended Use: Ideal if you know you will have a regular need for optimization and want to maintain a constant stock of credits.

2. On-Demand Credits (Top-up)

They can be added at any time for a defined period, offering maximum flexibility.

• Recommended Use: Perfect for running a test phase, for one-off needs, or if you have used up your monthly credits and want to push one last optimization without waiting for renewal.

Credit Estimation and Tracking in the Interface

To help you manage your consumption, Feed AI estimates the number of credits required before you even launch a task.

• At the "Case" level: Just below the number of products impacted by your rule, you will see an estimate: "approximate credits needed / remaining credits".
• At the "Task" level: Next to the "Launch" button, a similar estimate ("estimated credit cost / total available") tells you if you have enough credits for the entire task.
• Important: If your available credit balance (the exact amount) is insufficient for the operation, the launch will be blocked.

You can also track your detailed credit consumption in your client monitoring area. There you will find consumption graphs and a clear visualization of credits used versus what you have left.

Credit Consumption Guide

Text Tasks (Estimated per 1,000 products):

• GPT-6 Luna (Recommended default): ~2.0 credits (~25% cheaper than legacy GPT-4)
• Gemini 3.1 Flash Lite (Google EU): ~6.2 credits
• Gemini 3.5 Flash Lite (Google EU): ~9.5 credits

Image Tasks (Estimated per 1,000 square 1024×1024 images):

• Flare / Sunburst (Low): ~105 credits | Medium: ~155 credits | High: ~420 credits
• Sunburst only (Very High): ~690 credits | Maximum: ~1,470 credits
• Nano Banana 2 (Google EU) (Low): ~225 credits | Medium: ~340 credits | High: ~505 credits

The Free Trial Path

The good news? The entire configuration process we will describe below, including the crucial preview step, is completely free. You can set up, test, and refine your instructions as much as you need without using a single credit.

This allows you to prepare your optimization with peace of mind before deciding to launch it and use your credits.

Configuring Your First AI Task: A Step-by-Step Example

To illustrate the process from start to finish, we will show you a concrete example: optimizing product titles. Keep in mind that this is just one of many possible applications. The true power of Feed AI lies in its flexibility, allowing you to build any data transformation scenario you can imagine.

Part 1: Configure and Preview Your Task (Free for All Licenses)

This first part covers all the steps needed to set up your instructions and check the AI's work. These features are available to all users.

Step 1: Creating a New Task and General Settings
Navigate to Data Transform > Feed AI and click New Task.
Configure your general settings:

• Choose Input for Task: Select the columns the AI needs to read (e.g. title, description, brand, color). You can select source imports and General formulas (channel formulas cannot be used).
• Explore Input Data: Click Explore to preview your data and verify that your columns are not empty before running the AI.
• Unique Identifier Field: Select your product ID (e.g. id or sku). SmartFeeds uses this ID to merge the new AI results back into your catalog.
• Sort Products (Optional): Order products by a specific metric (e.g. sort by lowest impressions to optimize your worst-performing products first).
• Row limit: The maximum number of products to process per run. Keep it small (e.g. 20) during testing to save time and credits.
• Temperature: Controls AI creativity (from 0 to 1):
  • Low (0.1 to 0.3): For precise and factual tasks (extracting attributes, data formatting).
  • High (0.7 to 0.9): For creative tasks (marketing descriptions, new title ideas).
  • Tip: If unsure, start low (0.2).
• Task Type: Choose between Text (titles, descriptions, attributes) or Image (packshots, background removal, lifestyle staging).
• Model: Select your AI model:
  • Text default: GPT-6 Luna (recommended: ~25% cheaper on credits, sharper semantic understanding).
  • EU Data Residency: Gemini 3.1 Flash Lite or Gemini 3.5 Flash Lite (Google EU).
  • Image generation: GPT Image 2.5 - Flare (fast, default), GPT Image 2.5 - Sunburst (high quality), or Nano Banana 2 (Google EU).

💡 Looking for a detailed comparison of all AI models?
Check out our dedicated guide: Feed AI - Choosing the Right Model to compare credit consumption benchmarks, EU data sovereignty options (including Mistral Small 3.1), and image capabilities.

Step 2: Configuring a Case

Click + New Case to define your instructions. A Case has three parts:

1. Select products:
   • All products: Keep the filter toggle OFF.
   • Specific products: Turn the toggle ON and add your condition (e.g. character_count(title) < 50).
2. Prompt:
   Tell the AI what to do. You can use the Prompt Library to load ready-to-use templates (titles, descriptions, packshots, lifestyle scenes), or write your own:
   • Adapt the default prompt: Simply update the language, category, or rules in the text box.
   • Image Options: For image tasks, enable Transparent background (PNG cutout) and choose up to 3 variants per product.
   • Note: Type @ or start typing to get dynamic suggestions of your saved input columns.
3. Schema:
   Defines the output columns generated by the AI (written in YAML):
   • Default fields: The template generates improved_title (text) and reasoning_steps (bullet points explaining the AI's logic).
   • Customize columns: Rename fields (e.g. new_title or extracted_attributes) to match your needs.
   • Tip: The pre-built schemas in the Prompt Library are already formatted for you.

Step 3: Preview and Refine Your Results
This is your most important step for checking your work. Note: Running a preview is now required before launching an optimization to calculate an accurate credit cost estimate. Click the button to see what the AI has produced based on your instructions.

The preview button is a dropdown menu offering two highly useful testing modes:

• Iterate preview: This default option runs the preview on the same products you just viewed. This is ideal when you tweak a word in your prompt and want to see how it changes the exact same product.
• Shuffle products: This option runs the preview on a new, random selection of products matching your filter.

Use this preview to see if the AI understood you correctly. The interface is split into two parts: your input data on the left, and the AI's output on the right. To make it easier to check your work, you have a few tools at your disposal:

• Column Selector: Use the search box to show or hide specific columns. This is useful to compare a specific input (like the original title) with a specific output (like the new_title).
• Resizable View: You can drag the vertical bar in the middle to give more space to the input or output columns.
• Copy Preview: Use the "Copy preview" button to easily share your results with a colleague for a second opinion. Display Note: To keep the preview clean, very long text in the display table is visually cut at 1000 characters. However, clicking the "Copy preview" button will always copy the complete, uncut text.

If the results aren't right, you can go back and adjust your Prompt, Schema, or Temperature settings until the output is perfect.

Part 2: Activate and Use Your Results (Full Feature Access)

Once you are satisfied with your preview, the next steps involve launching the task at scale and using the data. These features are part of the full Feed AI functionality.

Ready to go further? The following steps, Launch and Download Results, are included in the full version of Feed AI. To unlock this functionality and apply your optimizations to your entire catalog, please contact our sales team

Step 4: Launching the AI Task

Once you are satisfied with your preview, make sure to save your configuration. Clicking "Save" creates a new Version of your settings. You can switch between different versions using the dropdown menu at the top.

New Action Bar: Action buttons and estimated credits are on the right (Launch, progress bar), while previous run metrics and product counts are on the left.

When you are ready, the launch button has two main modes:

• Launch X new rows: This is the standard option. Feed AI remembers which products have already been processed in the past. This option intelligently processes only the new or unprocessed products that meet your criteria. This saves time and prevents double-spending credits.
• Reset & Launch: By clicking the dropdown arrow, choose this option to restart the optimization from scratch. It will re-process all products that match your filter (up to the row limit), completely ignoring the history. You must use this option if you change your prompt and want those changes applied to products that were already optimized in a past run.

Step 5: Accessing and Using Your Results

Once a run is complete, the Download Results button becomes active. Here is how to manage and use your generated data:

• Review Past Runs: Use the execution selector dropdown at the top of the interface to navigate and review previous runs.
• Monitor Errors: If any products fail during the AI process, the interface displays the exact number of failed products at the end of the execution. This makes it easy to monitor your data health.
• Download & Merge: Click download to export a CSV file containing your newly generated data. This file automatically includes the unique identifier you selected. You can then import this file back into SmartFeeds and use a Join to merge it with your main product feed (Example: replacing the old titles with the new, AI-generated ones).
• Advanced Uses: Think of this downloaded file as a brand new data source. You can use it to extract specific attributes (like "material" or "style"), build powerful filters for your campaigns, or even create a custom quality score.

Quick Answers to Common Issues

Why can't I select a channel formula (e.g. google_shopping_title) as an input for Feed AI?

Feed AI only reads source imports and formulas created in the General section. To use channel logic as an AI input, move or recreate your formula in the General section first.

I updated my prompt or input columns. Should I click "Launch" or "Reset & Launch"?

Use Reset & Launch. This applies your new prompt to all products. The standard Launch button only processes new products that were never optimized before.

My filter is correct, but some products are not processed by my Case. Why?

Cases run in order from top to bottom. Once a product is processed by Case 1, it cannot enter Case 2. Place your most specific Case at the top of the list.

How can I ensure our product data stays strictly within Europe?

Select an EU-compliant model: Gemini 3.1 Flash Lite, Gemini 3.5 Flash Lite (Google EU), or Nano Banana 2 (for images). All prompt processing and data storage remain strictly inside the EU.

Related Articles

For further assistance with SmartFeeds, consider reviewing these articles:

• Join
• Mapping
• Filter

For additional help, feel free to reach out via our Contact Us page.