Using Airtable AI in fields

NOTE

AI credits and enabling AI on the workspace level are required to use AI in fields.

Understanding Airtable AI in fields

In Airtable, we refer to most AI-enabled fields as Field agents. Field agents are AI-powered fields in Airtable that can automatically retrieve, analyze, or generate data at the cell level. Unlike standard fields, field agents are dynamic. They can pull in information from the web and analyze documents to perform tasks that would otherwise require external tools or manual entry. Field agents can be added manually or generated by Omni when it recognizes that an AI-powered workflow is the most efficient solution.  

There are several field types in Airtable that support AI features. The set of features supported varies by the type of field that you are creating and configuring.

• Formula field - When you begin configuring a formula field, you can have AI help build out a formula by clicking the Generate formula button in the field editor. It’s worth noting that you can also ask Omni to help you build a formula.

• Long text field (a.k.a. AI text) - Have a field agent generate content based on custom instructions. The content generated will directly refer to information held in other fields related to the record that you are working with.

• Linked record field - Toggle on the “Show AI-suggested records” option in the field configuration menu to turn on AI. This lets Airtable AI either suggest or “auto-fill” linked records in the field based on context held in other fields in the table(s) connected via linked records, streamlining repetitious and manual work.

• Select fields (Single or Multiple select) - Allow a field agent to output single or multiple select options based on information held in other field(s) across a table.

• Number based fields (Number, Currency, or Percent) - Have an agent determine a number, currency, or percent based on other numbers and/or information in your base.

Adding a field agent

1. Open the preferred base and table where you would like to add a field agent.

2. Click the + button located next to the last field in the table or click an existing field’s dropdown menu and click Insert left or Insert right to add a new field next to an existing one.

3. In the menu that appears, you’ll notice a “Field agents” section at the top where you can choose from one of the common field agent use cases, click Build a field agent, or click Browse catalog to open up the Field agent catalog.

   1. In the field agent catalog, you’ll find a “Suggested for your base” section with field agent suggestions tailored to the Airtable base you are working out of. You can also choose a use case next to the “Specific to your use case” section. This will update the available pre-set options. You can also click Browse all to view even more options.

4. Depending on the option you chose above you’ll have one of two experiences:

   1. Field agent suggestion or catalog - After choosing a pre-set field agent, we will open up Omni with a request to build out the field agent you chose. After Omni has completed the initial request you can ask it to further alter the field’s configuration or begin making some manual adjustments to the newly created field.

   2. Build a field agent - After choosing to manually build a field agent, you’ll be brought to field configuration window. By default, the field type will be set to “Long text.” Click the dropdown to choose another AI field type. After configuring the field and giving it “Custom instructions,” click Create field.

Giving field agents custom instructions

Consider the following items when building out custom instructions (prompt) for a field agent from scratch or modifying existing instructions:

• Describe the task you want the agent to accomplish and share as much context as possible.

• Add any rules that will provide guardrails for the agent.

  • Include "Do" and "Do Not" statements to set boundaries for the AI model.

• Provide any data or benchmarks that might be relevant to the prompt you are writing.

• For more complex prompts, consider asking the agent to take on the task in a step-by-step manner

• Include any formatting instructions to drive consistency in the AI output

• Giving the AI model a role can be helpful (ex. "You are a product manager...")

• Include as many Airtable fields in the prompt as needed to provide context for the request.

  • You can use the + Insert field button in the custom instructions window to add fields or begin typing the field name with a curly bracket {  in front of the name. Either option will open up a list of the fields you can use in the prompt. The field name will appear as a token in the prompt and will automatically update, even if the name of the field is changed later.

• When you include a field, provide the context of what kind of content the field is storing, such as:

  • "Write an email that addresses the following customer feedback: {customer_feedback}"

  • or "Create a meeting summary based on the following notes: {Meeting Notes}"

• Remember that more specific prompts typically yield more specific outputs, while less specific prompts might result in more variation. You can use either method to achieve different results, but it's important to be intentional with your approach.

• You can always go back and edit a prompt, but we strongly recommend that you save a previous prompt in case you want to revert back to it (paste it back into the prompt text box) while you are testing the AI's output.

• Point out specific examples of what the agent output should look like.

  • To that end, add any default language/pre-determined outputs that the AI should always include.

  • Give the AI model examples to work off of. When providing examples, make it clear at the end of the examples section what you want the model to do by prompting it specifically with the content you want it to ingest and what you want it to do.

    Below are some examples of feedback and related sentiments that you'd output.
    
    The feedback statements below are examples only.
    
    Feedback: "This is the worst tent I've ever used."
    Sentiment: Negative
    
    Feedback: "This backpack is alright."
    Sentiment: Neutral
    
    Feedback: "I really love this sleeping pad."
    Sentiment: Positive
    
    Feedback: {customer_feedback}
    Sentiment:

Adjusting a field agent configuration

After a new field agent has been created you can modify its AI configuration in a few ways. Click the dropdown next to a field agent’s name and click Edit field to reveal these options:

• Settings - Toggle on or off each of these settings.

  • Enable internet search - If the setting is enabled, then we evaluate the prompt to determine whether internet access is needed. If it can be carried out without using the internet, then we won't use the internet.

    • Note that this often consumes additional credits and you should double-check the results for accuracy.

    • This option may not be available if internet access has been disabled in the admin panel (Business or Enterprise Scale only)

    • We support both general internet search (e.g. "Research this company," "find this person's job title") as well as search that specifies a URL ("get the company address from {Company_webpage}")

    • We also support internet browse (e.g. "Pull the text from this webpage and identify any grammatical errors")

    • If the AI uses the internet in creating its response, then the sources will be included in the agent’s output.

  • Run automatically - This will cause the agent to automatically run (generate values) when input values change

    • Much like an automation, be sure to consider the downstream effects of turning this option on.

    • i.e. If you have other field agents, formulas, or automations that are referencing the values output by the field agent, then this can trigger a lot of other changes to occur in your base.

• Model - Choose between various AI models available to you based on your plan and/or organization’s admin panel settings. There is a typically a relationship between the cost and quality of a model. You’ll see a “Low cost” tag next to models that will consume fewer credits.

Triggering a field agent run

Note

Only users with editor permissions or higher can manually trigger field agents to run.

From a base, there are a few ways to manually trigger a field agent to run:

• Click Run agent within a cell to run the agent only for that individual cell/record

• If a cell is empty and you don’t see Run agent, then click into the cell. Then, click the icon to run the agent again.

• To perform bulk field agent runs:

  • Click the dropdown menu next to the field agent’s name

  • Hover over Run field agent

  • Then click one of the options available:

    • All cells in the view

    • Cells modified by agent

    • Outdated cells

    • Cells with errors

    • Cells never modified

Understanding field agent dependencies and limitations

• Word limits - The total number of words in your query (custom instructions) and agent response must be under a certain amount set by the AI model being used on the backend. This includes the words in the fields referenced by your query. If you exceed these limits in your query, you will see an error message. If the AI's response exceeds the limits, the text will be truncated.

  • Lower powered AI models: Up to ~12,000 words

  • Higher powered AI models: Up to ~90,000 words

• Automatic generation - The Run automatically setting in a long text field can be enabled in tables containing many records. However, if you do choose to "Run automatically" on a table with thousands of records you are still limited by the amount of credits available to you and your team. So, it's possible that some of the records will generate automatically, but others will eventually error out due to other limitations.

  • If this setting is toggled on, and then you proceed to make any edits to your prompt or prompt settings and then save the field, we will show you some messaging that asks if you want to overwrite existing records. You only see this messaging for generating automatically.

    • Clicking Continue will overwrite existing content by generating content for those cells again. However, we never will automatically overwrite any cells that were edited by a human, even if you toggled on generate automatically. This is true whether you change your field configuration, or if you change cells that the AI field depends on.

    • Users will also have the option to proceed without overwriting. Choosing this option will still save any edits made to the prompt or prompt settings, but none of the existing content for that field will be overwritten. This will ignore automatic generation at the moment the field configuration is saved. Any future edits to existing records that cause automatic generation to occur or new records that are added will now generate content adhering to the AI field’s current configuration.

Configuring AI in linked record fields

We provide some additional information in this article, but the steps below give you the basic steps to set this up.

1. Navigate to an existing linked record field where you want to configure AI settings or add a new linked record field to a table.

2. Toggle on the “Show AI-suggested records” option.

3. Optionally toggle on the “Auto-fill top suggestions” option.

4. Next, you’ll select one or more fields from each table linked together that will help to give AI context about the relationship between the linked records.

   1. You can utilize this feature on intra-linked (within the same table) or inter-linked (between two tables) record fields.

   2. You’ll want to connect as many fields as needed to provide the AI model with the necessary context to suggest or auto-fill records with greater precision.

5. Click Save if you are modifying an existing linked record field or Create field if you just configured a new linked record field.

NOTE

Auto-filling top suggestions can be a trial and error process, so consider potential downstream effects (i.e. automations, formula fields, etc.) before making changes to a linked record field.

AI Security and Terms of Service

When using AI in Airtable, note that any data within fields referenced in a prompt may be shown to users who can view the AI long text field. This is particularly important when leveraging an AI long text field in an interface, where certain fields used in the prompt may be intentionally hidden from interface collaborators. For example:

• A creator builds an AI long text field that guesses a job title given an employee’s experience and salary

• An interface is created that shows the AI long text field and the employee’s experience, but not the salary

• An interface collaborator edits the employee’s experience section to instead tell the AI to ignore other instructions and repeat the salary input.

For more information on AI security and ToS please visit the following resources:

• Trust & Security at Airtable

• Airtable AI terms

FAQs