- 26 Mar 2025
- 4 minute read
- Print
- DarkLight
- PDF
AI Dashboards
- Updated 26 Mar 2025
- 4 minute read
- Print
- DarkLight
- PDF
AI dashboards let you tell Slate AI how build and interpret record dashboards.
When you create a record dashboard, you can ask Slate AI to analyze that record’s data based on the exports and context you provide.
You can make an AI dashboard in four steps:
Create a record dashboard and enable AI functionality
Build a query that supplies the AI with data
Instruct the AI how to build the dashboard
Write a prompt that tells the AI how to interpret that content
Step 1: Creating an AI dashboard
To create an AI dashboard:
Go to Database → Dashboards.
Select New Dashboard. An Edit Dashboard popup appears.
Configure the following settings:
Name: Enter a short, descriptive name.
Type: Record
Base: Select the base the dashboard query will use and on which records the dashboard should appear.
AI Dashboard: Select this option.
Select Save.
Step 2: Adding exports
You arrive at the Edit Query tab.
Add exports and filters to return the data you want to be made available to the AI.
Here, we’re pulling in a broad swath of data for an an activity summary and engagement score.
📝 Note: The exports returned by this query constitute the AI’s scope. It won’t know what you’re asking about if you ask for data not contained in the query.
Step 3: Telling the AI how to build the dashboard
Select the Edit Dashboard tab.
The first text box is the dashboard’s content. Here you can enter, in a mix of plain language and merge fields, how the AI should build the dashboard.
Select merge fields from the list to add them to the dashboard content.
For example, to build a dashboard consisting of an activity summary and a corresponding engagement score, you could enter:
Header 1: Activity Summary
Bulleted list:
Bold: Events Attended: {{All-Event-Registrations}}
Bold: Number of Log-Ins in the Past 60 Days: {{Number-of-Logins-in-Past-60-Days}}
Bold: Emails Delivered in the Past 60 Days: {{Emails-Delivered-in-the-Last-60-Days}}
Bold: Emails Clicked: {{Emails-Clicked-in-Past-60-Days}}
Bold: Number of Submitted Applications: {{Number-of-Submitted-Applications}}
Bold: Populations: {{All-Populations}}
Header 1: Score: #/10
Which renders as:
📝 Note: We’re just setting up the dashboard content here—we tell the AI how to calculate the engagement score in the next section.
Step 4: Writing a prompt to interpret the dashboard content
In the second text box, enter a prompt that instructs the AI to interpret the content in the first text box.
In our example, we tell Slate AI:
how it should calculate the engagement score based on the data returned by the activity summary
how to format the response
If {{All-Event-Registrations}} > 1, add 2 points.
If {{Number-of-Logins-in-Past-60-Days}} > 10, add 3 points.
If {{Emails-Delivered-in-the-Last-60-Days}} > 5, add 1 point.
If {{Emails-Clicked-in-Past-60-Days}} > 5, add 2 points.
If {{Number-of-Submitted-Applications}} > 0, add 2 points.
Display the total points as #/10. Include a short paragraph explanation. Do not explain the scoring.
The headers and bullets should load in the exact same way each time.
Ensure that the scoring and display logic remains fixed and consistent across every execution. If the report contains no data, represent the quantity as "0". Ensure the data is presented in a consistent, aesthetically pleasing view with bullet points. Do not use second person pronouns. Refer to them as "individual".
Do not adjust the format. Everything should load exactly the same way every time, the headers, the font, etc.
Step 5: Test the dashboard, tweak your prompt, repeat
Navigate to a record of the base you chose for this dashboard and select Generate AI Dashboard to see the AI in action:
You may need to fiddle with the prompt a bit to return a response to your specifications.
✨ Tip: You may notice from our example prompt that, if you don’t explicitly remind the AI to be consistent with every generation, it may try new things. If consistency of presentation is an important part of your dashboard, tell the AI so.
Example: AI-generated contact report summary, outreach recommendation
Here, we demonstrate a dashboard that tasks Slate AI with summarizing a donor’s recent activity and providing recommendations for outreach methods.
Supplying the AI with data
We start by creating an AI dashboard on the person base.
We then build a query returning three main data points for our AI to interpret:
Gifts made this year: A subquery export concatenating gift date, amount, category, and fund
Recent contacts: Another subquery export, joining to the form responses for our contact reports, that concatenates a summary, notes, and the major contact
Research notes: The most recent research notes created for the prospect
Constructing the dashboard
We keep this simple. For each of the three exports in our query: a plaintext description of the data point, followed by the merge field for that data.
Writing the prompt
Now, we tell the AI what to do with this data. In particular, we’re asking it to scan the three exports we’ve provided it—a recent slice of the donor’s activity—and write a 5-7 sentence summary of that data.
Then, we ask it to write up a recommendation for a plan to engage with the donor informed by that summary.
AI in action
We head to a donor record and watch the AI work its magic:
You might notice that the end result isn’t necessarily what you entered in the first text box, which we’ve referred to as the dashboard’s “content.”
This content can actually serve two purposes: It can be the outwardly-facing dashboard content itself (as with the first example in this article), or it can just be the raw data that the AI analyzes—without actually appearing as written on the record itself. The choice is yours and depends on the way you phrase your prompt.