You can use models from OpenAI, Anthropic or xAi to analyze and modify data using AI. These models can perform a huge variety of tasks, including translation into most common languages. You would simply provide it with the appropriate prompt and it would perform the action (e.g. "Translate this text to French", "Summarize this text", "Rate this applicant on a scale from 1 to 5 based on their past experience")
OpenAI ChatGPT
Our Analyze and Modify Airtable Data with ChatGPT automation uses OpenAI's GPT models.
This automation also supports attachments.
Configuration
You'd need to add an OpenAI API Key first, as explained here.
Navigate to your dashboard and create the automation in the right Airtable base.
Under Setup, set the OpenAI API Key you added in step 1.
Set the System Instructions. You can provide context to the model here for the task at hand.
Set the Prompt Source to either Static Prompt or Airtable Field. A static prompt would be identical for every record, whereas a prompt from an Airtable field would be different for every record (containing data from that record, e.g. for analysis).
Set the Prompt to something like "Translate this text to French" or pick the Prompt Field.
Then set and Output Field.
Optional: Add Attachment Fields to pull in files for processing.
Under GPT Model, choose the model to use
Optional: Define parameters for the model under Advanced
Anthropic Claude
You can use our Analyze and Modify Airtable Data using Claude automation to send the contents of an Airtable field to Anthropic's Claude AI model and save the response back to your records. Here's how:
Select the Airtable Table containing the records you want to process.
Optional but recommended: Add your Anthropic API Key in workspace settings. Without your own key, only test mode is available (limited monthly character allowance per workspace).
Select the Input Field β the field that contains the text you want to send to Claude.
Optional: Add an Additional Prompt that gets appended to the input. You can choose between a Static Prompt (the same text for every record) or an Airtable Prompt Field (a different prompt per record).
Select the Output Field β the field where Claude's response will be stored.
Optional: Enable the Error Message Field and select a text field where the automation will log error messages for records that encounter processing issues.
Choose a Model: Claude Sonnet 4 (default), Claude Haiku 4.5, Claude Opus 4, or Custom Model (specify any Anthropic model name).
Optional β Advanced settings:
Output Max Tokens β the maximum number of tokens in the response (default: 4096; up to 8192 with your own API key).
Output Temperature β controls creativity from 0 to 1 (default: 0.7).
Output Top P β probability threshold for top-p sampling, 0 to 1 (default: 1.0).
System Message β system-level instructions for the model (max 1000 characters).
Without your own Anthropic API key, the input plus additional prompt is limited to 3,000 characters per record, and there's a monthly character allowance per workspace. Adding your own API key removes both limits (you'll be billed directly by Anthropic).
xAi Grok
The Analyze and Modify Airtable Data using the Grok Model allows you to use the xAI Grok model to process the data in your Airtable fields. This is an easily set-up automation that can be a powerful ally in understanding your data.
To process more than one record at a time, you'll need an xAI API Key. Follow the steps outlined in this article to add one to your workspace.
π‘ Note: You can test the automation without an API key in test mode.
Setup parameters
Table selector: Choose the table containing the data you want to use. You can change this selection as needed.
xAI API Key selector: Select the API key.
Input field: Choose the field that contains the data you want to analyze or transform.
Additional Prompt Type (Optional): Set a static or dynamic prompt.
Additional Prompt (Optional): If a static prompt is selected, enter it here; if using a dynamic prompt, select the Airtable field with the prompt for each record.
Output Field: Choose the field where you want to store the result after processing the input field.
Processing Behavior
Skip: Records with an already populated output field wonβt be processed.
Overwrite: All records will be processed, even if the output field is already populated.
Advanced
In the advanced section, you'll have access to advanced settings such as:
Output Max Tokens: The maximum number of tokens to generate in the response. The max allowed is 4000 and requires using your own xAI API key. Each token corresponds to ~= 4 characters (This is ignored if a personal API key is provided).
Output Temperature: The temperature of the model. Higher values will result in more creative responses, but also more mistakes.
Output Top P: The probability threshold for top-p sampling.
System Message: You can provide a system message to the model here.
In the run section, you can copy a formula to process one record at a time, run the automation as a script inside Airtable, schedule it to run every 15 minutes, 1 hour, or 24 hours, or trigger it using a Webhook.