AI FAQ
➡️ What's the difference between the AI Assistant and the Dashboard Q&A?
The Dashboard Q&A works within the context of the specific dashboard you're viewing. It uses its associated metadata—queries, calculations, lineage, pinned assets, and more—to generate answers.
The AI Assistant (also referred to as the AI Search) draws from a broader set of relevant assets across the catalog, not limited to one dashboard. It uses metadata and content from the most relevant assets to respond to your question. To be more specific, the AI Assistant can:
Handle a conversation - taking into account multiple messages to adapt its answer
Search on different assets types : tables, columns, dashboards, knowledge pages and their pinned assets
Use assets metadata to create an answer and provide its sources
Ask for clarification if theres is a doubt on action to take or if question is out of its scope search
A summarized comparison between the two is available here.
➡️ AI Assistant - What information is being used?
Curious about what exactly powers the AI Assistant’s answers? Or looking for ways to boost its performance by strengthening your documentation and enriching your assets metadata? Check out the full AI Assistant Context page.
➡️ Dashboard Q&A - What information is being used?
Wondering how Dashboard Q&A knows what to answer? It relies on your dashboard’s metadata, such as titles, descriptions, fields, and linked datasets, to provide contextually relevant responses. You can find the full list of metadata used by the Dashboard Q&A here.
➡️ Describe with AI - What information is being used?
The Describe with AI button generates an asset description based on its contextual information (SQL source query and all other available metadata). You can find the complete list of elements and metadata used by AI to suggest a description here.
➡️ What type of questions work well today?
AI Assistant (AI Search) is best for questions based on asset definitions and related pinned content. Examples:
“What tables can help me analyze
<concept>?”→ Uses table descriptions and Knowledge page pins related to the concept.
“In which dashboards can I track
<metric>?”→ Finds dashboards connected to relevant data assets.
“What is the definition of
<term>?”→ Pulls from documented definitions and descriptions.
“What tables contain the column
<column_name>?”→ Uses lexical search (not semantic), based on keyword matches.
🎯 Tips:
Specify the asset type in your question when possible.
“Do we have a knowledge page about
<concept>?”“Do we have a dashboard analyzing
<concept>?”
Use terms that most probably appear in asset descriptions or titles.
Dashboard Q&A is best when you need insights within the scope of a specific dashboard, such as:
“What is this dashboard about?”
How is this metric calculated?”
“What filters are applied here?”
“Which assets power this dashboard?”
➡️ Best practices to improve AI accuracy and surface key assets
To ensure better results and increase the visibility of your most important data assets:
Write rich, clear descriptions
Add context and FAQs directly into asset descriptions—especially for frequently asked questions by stakeholders. Prioritise assets descriptions for the ones that you’d like to see being easily researchable.
Certify/deprecate assets
Certified assets are boosted by the AI assistant, whereas deprecated assets are strongly penalized and have less chance to be part of the Assistant’s answers. Indeed, certification boosts the asset’s score by 1.5×, whereas deprecation cuts the asset’s score in half, making it very unlikely to appear in results
By using certifications, you guide users by clearly marking trusted or outdated content.
Pin relevant assets to Knowledge pages to make important relationships explicit and improve discoverability.
Use consistent naming and tagging to ensure metadata is clean and searchable.
➡️ Where can I leverage the AI assistant?
Through the Catalog app
On Slack
On Microsoft Teams
On the Chrome extension
The AI assistant used on the Chrome extension is also called the Dashboard Q&A, and its scope is narrowed on a specific dashboard.
OnDust
Through our public API
➡️ What's on the AI Assistant roadmap?
We’re actively expanding AI capabilities. Upcoming improvements include:
Support for lineage-based queries
Enhanced use of additional metadata such as owners, tags, and more
Last updated
Was this helpful?