Workflows: AI Workflow Node

This article explains the AI Workflow Node, which can be configured as part of Workflows to perform AI-driven tasks, such as generating content, analyzing images, or performing calculations based on given instructions. Learn more about using workflows in Clappia here.

Help Video

Used For

• Automating content generation, summaries, or calculations within workflows.
• Extracting information or analyzing images, such as detecting defects in uploaded photos.
• Creating dynamic responses or outputs that can be used in subsequent workflow actions, like emails, notifications, or submissions.

Editing the Node

After you have added the AI workflow node to the app, you will be able to configure it. Once you have added the node or clicked on it, a right panel will appear where you can proceed to configure that node.

‍

Step Name

This is the name that appears in the Workflow on the left side (on the node). If there are multiple AI Workflow Nodes in the workflow, set different step names like AI Analysis for Defects or AI Text Generation.

LLM

Select the LLM provider to be used for the AI step. Available options include Claude, Gemini, and OpenAI.

Each provider generates slightly different responses. You can experiment to see which one works best for your use case.

AI Model

Once an LLM is selected, this dropdown shows the corresponding list of available models. Choose a model from the list to power your AI interaction. The different models available allow you to experiment and find the one that provides the most accurate or relevant responses for your task.

Note: The output may vary slightly depending on the model chosen, even if the instructions are the same. If needed, test your instructions on the provider’s platform before using them in the app to reduce unnecessary AI calls.

‍

Instructions

Type specific instructions here for the AI to follow. Type in the exact instructions to display the data that is needed and to remove any unwanted text generated by the AI. For example:

• “Summarize the inspection report uploaded.”
• “Detect any defects in the machinery control panel image and generate a list of issues found.”
• “Generate a response email to inform the applicant of their application status.”

You can use field variables from the app to dynamically include data in the AI's instructions, enabling personalized and context-aware responses. To pull variables from the app, simply type in @ followed by the field name.

Note: Tweak your instructions to get the required results before distributing the app to users.

File Handling

The AI Workflow Node supports analysing uploaded PDFs and images via field variables. For example, you can use variables from a Camera, Images & Files block to allow the AI to interpret the file contents.

Note: Claude can read PDFs of up to 100 pages. Only document file types are currently supported. More formats may be supported in future based on LLM capabilities.

Parent Step

The parent step refers to the function that is directly above the AI Workflow Node in the workflow. This might be a Start node or another preceding workflow node.

Enable This Step

This step is enabled by default. Disable this step to stop the AI Workflow Node and any nodes below it from executing.

Variable Name

Once the AI Workflow Node is saved, it generates a unique Variable Name for the node (e.g., ai). This variable stores the output generated by the AI, based on the instructions given. You can use this variable in subsequent workflow nodes to reference the AI-generated output directly, allowing for dynamic automation based on the AI results. Again, to pull up the variable for the workflow node, type in @ followed by the workflow node’s name.

Using Variable Name in Other Workflow Nodes

The Variable Name created by the AI node can be integrated into other workflow nodes to automate actions. Here are some examples:

• Email Node: Use the variable (e.g., {ai}) in the Body field of an Email Node to automatically include the AI-generated output in emails. For example, if the AI generates a response email to a customer inquiry, the variable can populate the body of the email, making it easy to send personalized responses without manual entry.
• If Node: The variable can be referenced in conditions within an If Node to trigger different actions. For instance, if the AI output is a summary of survey feedback, you could use LEFT({ai}, 10) = "Positive" to check if the summary starts with “Positive,” triggering different follow-up actions based on sentiment.
• Create Submission Node: You can use the variable to automatically populate fields in a new submission, creating a record of the AI’s output within another app.

This variable-driven approach enhances flexibility, enabling you to apply the AI output dynamically across different parts of the workflow.

Example

A manufacturing facility performs routine inspections on control panels for critical machinery to ensure ongoing functionality and safety compliance. Inspectors photograph the panel, and the AI Workflow Node analyzes the image to identify potential issues due to wear and tear.

Workflow Configurations

AI Node
Step Name: AI

LLM: OpenAI

AI Model: gpt-4o

Instructions: Analyse the image in {upload_ima} to check for defects such as missing buttons, damaged displays, loose wiring, misaligned labels, Loose or Missing Screws and Rust or Corrosion. Generate a report that begins with 'Defect Detected' if issues are found, listing each defect. If no issues, return output as 'No Defects'.

In the instructions, the variable {upload_ima} has been used as there is a Camera, Images & Files block added to the app with that variable name.

If Node

Condition: LEFT({ai1}, 15) = "Defect Detected"

If Node

Condition: LEFT({ai1}, 15) = "No Defects"

Email Nodes

Body: {ai1}
Rest of the configurations for this node as per user requirement. Check here to see how to configure the email node.

Once a submission is made with the image uploaded;

The workflow begins, and the AI node processes the image as per instruction. As per the workflow the email is received as seen in the image below;

The body of the email, instead of being manually written, dynamically takes the output generated by the ai which is a complete report as asked in the instruction of the AI node. Modify the instructions to get different results as per requirements.

‍
Usage Limits and Add-on

The AI Workflow Node has a monthly usage limit. It shares this limit with the AI Block used in the Design App. If the total usage exceeds the limit, additional calls can be made through an AI Add-on.

To increase AI usage, you can go to Workplace Settings > Billings tab and select the add on. Your AI usage can also be seen in the Billings tab.

‍

Note:

• This is a beta version of the AI service available as an optional add-on. It is not included in any plan.
• The block uses APIs to interact with external LLMs. Use of this feature implies agreement with the terms and conditions of the selected model provider.
• The AI Workflow Node has a monthly output limit of 100. If you are also using the AI Block in the Design App, the combined output limit for both the AI Workflow Node and the AI Block will be capped at 100 per month.
• To avoid exceeding the usage limit in Clappia, you can test and refine your instructions directly on the AI platform. Once you’re satisfied with the responses, apply them within Clappia for final testing.