Kai Use Cases & Examples

Practical examples of what you can ask Kai to do.

Kai Capabilities

Troubleshooting

Debug a failed job:

"Job 789012 failed. Read the error message and identify the root cause."

Investigate data quality:

"I'm seeing unexpected nulls in the customer_revenue table. Help me trace where they're coming from."

Kai reads actual job logs, checks configurations, and traces data lineage to provide specific solutions.

Create a transformation:

"Create a customer lifetime value model using data from our CRM. First outline the approach, then build it."

Convert Python to SQL:

"Convert this Python transformation to SQL for better performance: [paste code]"

Complex analytics:

"Build a cohort analysis showing monthly customer retention rates over the past year."

Configure an extractor:

"Set up a Shopify extractor to pull orders, customers, and products."

Custom API integration:

"Create a generic extractor for the TikTok Ads API with pagination and OAuth authentication."

Add packages:

"Add the pandas package to my Python transformation environment."

Create a dashboard:

"Create a sales dashboard with a date picker and bar chart showing monthly revenue by product category."

Interactive analytics:

"Build a customer segmentation app that lets users adjust RFM parameters and see segments update in real-time."

Generate project docs:

"Generate documentation for the customer analytics flow, including all transformations and their business purpose."

Update metadata:

"Generate descriptions for all tables in the customer_data bucket."

Project overview:

"I'm new to this project. Give me an overview of the data structure and what each bucket contains."

Business queries:

"Calculate our top 10 customers by revenue for Q3, excluding refunds, and show year-over-year growth."

Build a pipeline:

"Build a complete pipeline from Shopify to Snowflake with data quality checks and customer segmentation."

Data migration:

"Help me migrate our legacy MySQL analytics to Keboola, preserving all existing business logic."

Instead of	Try
“Build a data warehouse”	Break into steps: “First design the dimensional model”, then “Create staging transformations”
“Calculate churn rate”	“Calculate monthly churn rate where churned = subscription ended and no renewal within 30 days”
One complex request	Iterate: basic version first, then add features incrementally

For more tips, see Best Practices.