Generate content with Airtable AI
Airtable
9 min read
Written by
Airtable
Filed Under
Airtable AI
TOPICS
Generate content in seconds with Airtable AI
With the Generate Content function you can create a first draft in seconds. Spend your time iterating, editing, and polishing your message to inspire audiences and compel them to act.
Content generation is one of the most time-saving applications of AI with Airtable. While leveraging AI has already proven to be a strong method to develop all sorts of content, from blog posts to marketing requirements documents, the pain of having to pull your data out of the tools you are already using, only to repeat that process after the content has been generated, is cumbersome and inefficient.
By embedding AI into your team’s existing workflows you can reduce the lift of generating content and bolster the impact of day-to-day work. In this guide, we’ll walk through the specific content generation features available in Airtable and share how your team can start using them today.
When would you generate content with AI?
Generate Content allows you to give AI instructions and any existing inputs on a piece of content that you need to create. You might want to turn an existing content brief for a marketing campaign into a blog post, landing page, social media posts, and more. Or you need to build a series of sales emails based on information on the companies you are targeting. Or you are a product manager who needs to take user-research and turn it into a product requirements document for your engineering team.
While the possibilities are endless, there are three scenarios that Generate Content is a particularly good fit for:
1) First drafts of documents
It can often feel impossible to get started on the first draft of a document. By leveraging AI for your first draft it can help overcome writer’s block and get the creative juices flowing. Even if you end up rewriting parts of the content, creating the first draft effortlessly can help you get to a finished product quicker than you otherwise would.
2) Automating repetitive content creation
In many jobs you are tasked with creating new pieces of content that are a derivative of the original but in a new format. For example, Marketers might create a blog post for a campaign but then need to create accompanying assets like social media posts, emails, and landing pages. Similarly, Product Managers write detailed Product Requirement Documents (PRDs) for new products they are building, but then have to deliver launch briefs, support documents, and a host of other deliverables that are essentially repurposed versions of the original document. By leveraging Generate Content after you have your initial piece of content developed, you can save time by editing AI drafts rather than spending time creating these accompanying pieces. Simply input the original piece of content, whether a document, video, or meeting transcript and generate content off of that.
3) Creating personalized content
Crafting personalized content is time-consuming, but often pays off. Whether you are addressing a sales prospect, a job candidate, or even an investor, personalized content increases core email metrics like open rates and click-through rates. Leveraging AI you can provide key details about the person you are communicating with (or even pull these in from another data source) and allow it to craft a message that’s personalized in seconds.
How to start generating content with Airtable AI
To generate content in Airtable use the AI field and Automations features. Here’s when to use each of these features.
Feature: AI field
When to use it: Generate content based on information in a single record using the AI field. The AI field can even automatically generate new content when a new record is created or info is updated in an existing one. Add generated content to a field in a base or interface.
Feature: AI in Automations
When to use it: Generate content based on information from one or more records as part of a multi-step workflow. You can choose to trigger the generation at a certain time or when an action is taken in an integrated 3rd party tool. You can also integrate information from multiple sources into a single content generation workflow to make the results more specific. Save the generated content to an Airtable record or send it to another service like email, Slack, or an external document.
Method 1: Use built-in templates
The easiest way to get started with generating content using the AI field is to use one of these templates:
Product > Product Requirements doc
Product > Product launch brief
Marketing > Marketing brief
Marketing > Campaign brief
Marketing > SEO keywords
Marketing > Tweet
UX Research > Research plan
UX Research > Interview guide
Recruiting > Job description
Recruiting > Outreach email
Recruiting > Search query
Recruiting > Candidate assessment
Each of these templates requires a different input. For example, you might input a long-form document and allow AI to provide condensed or re-written formats of the source material. Here are some examples of content you might generate after inputting long-form content:
Product launch brief - based on a product requirements doc (PRD)
Marketing brief - based on a PRD for a feature or product
UX Research interview guide - based on a research plan
Some of these templates have multiple inputs, such as the research plan and job description templates below. In these cases, you need to have fields that map to these specific inputs in order to use these templates. Some additional optional fields:
“Include additional fields”: a place to input any additional fields that you would like to be included in the AI prompt
“Add your own examples”: You can paste one or more examples of a good output in this field to push the LLM to write something similar.
“Give any other instructions”: You can choose one of our predefined styles, or define your own.
The templates are a good place to start, but you may want to use custom prompts if:
You want to generate content that is not any of the types included in templates.
You have a specific tone or voice that you want the LLM to use.
You have detailed instructions you want to be followed or a complex ask.
Method 2: Use custom prompts
If you’ve started with a template and then want to move onto a custom prompt to build a more flexible AI query, you can convert your templated query to a custom one by clicking the “change to a text box” button. This will allow you to see the prompt that was created based on your configuration in the template wizard.
Here’s an example of a custom prompt based on the job description template, but with a few elements that are going to make it better at producing what you want. In this example we added more detailed up front context and more control over where the examples go in the prompt.
You can also play around with the settings in the prompt configuration to get the LLM to tweak the outputs:
Model: At launch, this will be a toggle that says “Use AI models that optimize for best results” and it will be defaulted to ON, which will use GPT-4 or Claude v2.
GPT-3.5 and Claude v1 are cheaper and faster, but GPT-4 and Claude v2 may be better at writing and following directions.
If you need to input one or more long documents or output something long, then “optimizing for best results” is the way to go. GPT-4 and Claude v2 have the ability to process much more data than other models.
Temperature: This refers to how “creative” the model is. The higher the number the more creative the output is, so a high number may be appropriate for idea generation for a marketing campaign, while a low number may be appropriate for generating responses to a customer email. If this is zero, the model will generate the same or substantially-similar content every time. If this is high, then the same prompt would result in different answers if run multiple times.
Generate automatically: If set, this will automatically generate the field when new records are created, and when inputs change.
A word on examples
Examples can be incredibly effective at getting the LLM to return something that is very similar in format and tone to the example, often surprisingly so. Often customers will talk about how they believe they need a model that’s trained on their specific data, when just sending a few examples of the output alongside the prompt can get them pretty close.
Method 3: Generate content with a single field
You may need to generate many different types of deliverables off of the same asset, and creating many different AI fields to do so is painful. In the example below, we use a single AI field to generate website copy, email copy, a social media post, and a blog post off of the same marketing brief. This is accomplished by:
Storing the marketing brief into a column, so it can be accessed by each row in the table.
Creating another table with Deliverable types, where we define the prompts that will be used for each type. For example:
Email: “You are a Marketing Manager working at a large wireless service provider and you need to write a promotional email for a campaign. In your response, please include an email subject line and the copy of the email. The campaign has the following description:”
When the deliverable type is chosen in the column “Deliverable type,” the field “GPT prompt” pulls in the language that will be part of the prompt.
The custom prompt that generates the website copy or email then just looks like this the screenshot below.
When the prompt is sent to the LLM, the field content will be inserted, and the prompt will end up looking like:
“You are a Marketing Manager working at a large wireless service provider and you need to write a promotional email for a campaign. In your response, please include an email subject line and the copy of the email. The campaign has the following description:
Campaign brief: This campaign will encourage families to sign up for our wireless service's family plan in advance of the start of the school year, as kids go off to school and will need to connect with their families. We should focus on families with both college and high school-aged kids…”
Method 4: Use AI in Automations
In Automations, you can create an action which makes a query to an LLM and gets back a response. You can then use that response in any of the ways you use Automations today - write an email, create a doc, post to Slack, write to a field in a base. In addition, you can input that response back to an LLM, creating the ability to chain LLM calls together.
One example of a content generation use case in Automations:
A team member completes a form describing a blog post that they’re going to write.
Draft of the blog post is generated and saved to a Google Doc.
Requestor receives an email with a link to their Google Doc.
Try it now
Generating content historically takes a lot of time and thought. With Airtable AI, you can generate content drafts in seconds. Spend your time iterating, editing, and polishing the right message to compel your audience to act. Whether you start with one of our pre-built templates or build your own prompt, try creating content with Airtable AI today.
About the author
Airtableis the digital operations platform that empowers people closest to the work to accelerate their most critical business processes. Across every industry, leading enterprises trust Airtable to power workflows in product operations, marketing operations, and more – all with the power of AI built-in. More than 500,000 organizations, including 60% of the Global 2000, rely on Airtable for digital operations and citizen development to help transform how work gets done.
Filed Under
Airtable AI
