# AI Access Image

The Assess Image stage uses AI to generate a summary or description of an image. It helps you understand the content or context of the image — such as what’s visible, the type of document, or key elements

![](https://531923687-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FnhvnNmu6p9CSODwQTjgf%2Fuploads%2F4k6rOuPl7Cj1GWvUk1yY%2F0.png?alt=media)

### **AI Assess Image**

**What is the Assess Image Stage?**

The Assess Image stage enables AI-powered image analysis within a workflow. It can:

* Detect objects
* Classify images
* Evaluate image content or quality
* Identify patterns or document types

**Important:**\
The Assess Image stage can only be used under a For Each stage, as it requires record-level context for accessing image fields.

**Why Use Assess Image Under For Each?**

When you have a list of image records (e.g., invoices, ID cards, documents) and need to analyze each image individually, place the Assess Image stage inside a For Each loop to:

* Loop through each record
* Access image fields (e.g., ${ForEachRepository:Image})
* Run AI analysis on each image
* Store the output for further processing

**Step-by-Step Configuration**

**Step 1: For Each Stage**

* **Source:** Select the repository that contains the image records (e.g., *Candidate Documents*).
* This will loop through each record in the repository.
* Fields from each record will be available using ${ForEachRepository:FieldName} syntax.

**Step 2: Add Assess Image Stage Inside For Each**

* **Field Selection:** Choose an attachment/image field from the For Each Repository Record (e.g., *ID Proof Image*).
* **Prompt:** Write a clear instruction to guide AI.

**Prompt Examples:**

* “Summarize what is visible in this image.”
* “Identify the type of document shown (e.g., ID card, invoice, receipt).”
* “Describe the background and objects in this image.”
* “Is this image a selfie with a document? Provide reasoning.”
* “Check if the image contains any official seal or stamp.”
* “Verify whether this Aadhaar card is front or back side.”
* “Check if the uploaded Aadhaar card contains a visible Aadhaar number.”

**Output Variable:** Store the AI result in a variable (e.g., Variable:AnalyseText).

**Step 3: Add Next Steps After Image Analysis**

You can now act based on the output:

* Add a Condition stage to evaluate the AI response.\
  Example:\
  Check if ${Variable:AnalyseText} contains "Government of India"
* Add an Update Record stage to store the analysis back into the repository.

**Output Example**

* Input Image: An uploaded Aadhaar card scan
* Prompt: “Verify whether this Aadhaar card is front or back side.”
* Output Variable: CardSideInfo