Ask your AI about your views
Questions that get your AI to describe your workspace as it is right now.
Your AI can describe the views in your workspace as they are right now: what's there, what each metric means, what's healthy and what was left out. These questions are the fastest way to learn your workspace, and the answers stay current because they come from your live views, not from a document.
How to read an answer
Every answer tells you where its number came from:
The view it used. For example, "answered from SV_CG_SALES_REPORT". That's the view behind the number.
What the metric counts. When your model defines a metric, the answer quotes that definition, so you can see what's included and excluded. When it doesn't, the AI describes the metric from its description and how it's calculated.
A label when the answer didn't come from a view. If no view can answer, the AI can query your Daasity data models directly, and as a last resort your raw connector data. It says so next to the number and names the table it used, so you know to check that number against your dashboards. A number from raw connector data is labeled as unreconciled.
What it held back. If your question touches customer-identifying fields (names, emails, phone numbers, street addresses or IP addresses), the answer lists which ones were withheld.
If none of your data can answer the question, the AI says it can't answer rather than guessing.
Find the right view
"List all of my views with their health."
"Which view should I use for net sales by channel?"
"Which views cover subscriptions?"
"Show me every view that has a product line field."
A keyword search only returns the best matches. If you're looking for a specific view that doesn't come up, ask for the full list.
Understand a metric
"What's the definition of Total Revenue, and which view is it in?"
"What does Net Sales include and exclude?"
"What default filters does the Orders view apply?"
Check health before you rely on a number
"What's the health of the view you just used, and how recent is its data?"
"Are any of the metrics in that answer ungoverned?"
"Is this view built on a derived table?"
Find out why something is missing
"What was left out of the Sales Report view, and why?"
"Why can't you answer that from Sales Report?"
"Which views were rebuilt today?"
Start from a saved Look
If there's a Look you already trust, give your AI its link or ID. The AI gets the Look's fields, filters and definitions and rebuilds the report from your views. It names the Look and the view it used, and lists anything it couldn't carry over, such as a field the view doesn't have or a relative date filter it had to interpret. This works with Looks saved in your shared folders.
"Reproduce this Look" followed by its link.
"Reproduce Look 1234, then break it out by channel."
Make it a habit
Ask your AI to name the view it used in every answer, and to say when an answer came from anywhere other than a view. Add that request to your standing context so you don't have to repeat it - see Give your AI standing context.
Last updated
Was this helpful?