diff --git a/modules/ROOT/pages/dashboards/ai-dashboards.adoc b/modules/ROOT/pages/dashboards/ai-dashboards.adoc
index 2d00f406b..4b2a03dc5 100644
--- a/modules/ROOT/pages/dashboards/ai-dashboards.adoc
+++ b/modules/ROOT/pages/dashboards/ai-dashboards.adoc
@@ -11,6 +11,9 @@ When creating a dashboard with AI, the AI analyzes your link:https://neo4j.com/d
 It cannot read the actual data in your database.
 ====
 
+The intelligent prompting happens in the background. 
+AI is going to analyse the schema, and it will try to make up some questions, and then it will convert each of these questions into a chart.
+
 If you do not enter a prompt, Neo4j AI will still make use of your database schema and come up with a suitable dashboard.
 
 Note that the following examples might differ when you reproduce them.
@@ -79,4 +82,14 @@ image::dashboards/ai-dashboard-dual-focus-1.png[]
 .A dashboard with both a data and a visualization focus (2)
 image::dashboards/ai-dashboard-dual-focus-2.png[]
 
+== Quality of the data model
+
+AI provides a great starting point, but the quality of your underlying data model still matters. 
+A graph model that has been thought out well, leads to a dashboard that tells more meaningful stories.
+AI infers nodes and relationships, but you might like to refine the model based on your specific questions - that way you will be able to reference entities directly from your schema (like `Customer`, `Order`, or `Category`) to guide AI towards more relevant charts.
+
+== AI as a starting point
+
+Remember, you can always edit an AI-generated dashboard.
+It's fun to use AI as a starting point, and then build on it by editing the output - and refining visualizations, colors, and layouts to match your needs.