In Atlas, some fields do more than just store data, they think. These AI-powered fields can search the web, perform calculations, and write content, cutting out the need for copy-pasting or switching between tools.
There are several field types in Atlas that support AI features:
- Text: AI can generate conent based on instructions you set, using data from other fields in the same record
- Number: AI can calculate or suggest number based on related values in the record
- Boolean: Let AI decide between true or false based on logic pulled from other fields
- Single-select / Multi-select: AI can pick one or more options based on context from surrounding data, helping standardise entries across records
- Date: AI can suggest or auto-fill a relevant date by understanding what the record is about
- Rating: Let AI assign a rating using your criteria, saving the need for manual review
- Phone / Email: AI can suggest or extract contact details based on info in other fields or linked records
Adding a AI field
- Open the Data table of a preferred dataset
- Click on the
+ Add field
button - Select one of the AI-supported field types above and toggle the
Use AI
switch - Toggle
Search internet
if you want to retrieve information from the internet - In the Instructions box that appears, write your custom prompt to the AI
- Validate the output by clicking on
Generate preview
- Then click on
Create field
Custom instructions tips
Consider the following items when writing custom instructions (prompt) for the AI:
- Describe the task you want the AI to accomplish and share as much context as possible
- Add any rules that will provide guardrails for the AI
- Include "Do" and "Do not" statements to set boundaries for the AI model
- Provide any data or benchmarks that might be relevant to the instructions you are writing
- For more complex instructions, consider asking the AI 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 environmental analyst")
- Include as many data fields in the instructions as needed to provide context for the request
- Use the curly brackets byttin to add fields
- Remember that more specific instructions typically yield more specific outputs, while less specific instructions might result in more variation.