KAI Assistant is a native AI assistant built directly into Kleene.ai. If you want to know where to start — this guide covers the six things you can do, with example prompts for each.
No setup required. No SQL needed. Just ask KAI Assistant.
Getting started
KAI Assistant is opt-in and role-aware. Once it's enabled on your plan, you'll find it inside the Kleene.ai platform. Access depends on your user role — if you're not seeing it, check with your workspace admin.
The chat interface works like you'd expect: type a question or request in the input bar and KAI Assistant responds inline. To help you get started, suggested prompts appear above the input bar — these are context-aware, so they'll reflect what's relevant to where you are in the platform.
Token usage is unlimited across all plans for the first three months.
1. Ask questions about Kleene.ai — get answers instantly
If you're new to Kleene.ai, or you've hit a feature you haven't used before, KAI can answer your questions directly from the documentation. No tab-switching, no Googling, no hunting through docs manually.
Try asking:
"How do I set up a new data source?"
"What's the difference between a transform and a group?"
"How does scheduled refresh work?"
KAI Assistant pulls answers from the Kleene.ai docs and returns contextual responses grounded in the actual documentation — not a generic chatbot reply.
2. Generate and optimize SQL transforms
This is one of the highest-value use cases for data engineers and analysts. Describe the logic you need, and KAI writes the SQL. Already have a query that needs improving? Paste it in and ask KAI to optimize it.
Try asking:
"Write a SQL transform that joins my orders table to my customers table on customer_id and filters for orders in the last 30 days."
"This transform is running slowly — can you suggest optimizations?"
"Rewrite this query to use a CTE instead of a subquery."
KAI Assistant understands your warehouse schema and metadata, so it can write transforms that are relevant to your actual data structure — not just generic examples.
3. Search and inspect transforms without navigating the UI
If you manage a large number of transforms and groups, finding the right one can take longer than it should. KAI Assistant lets you search by name, description, or what a transform does — in plain English.
Try asking:
"Show me all transforms in the marketing group."
"Find the transform that calculates customer lifetime value."
"What transforms are currently failing?"
Once you've found what you're looking for, you can inspect its contents, view the SQL, and understand what it does — all from the same KAI Assistant conversation.
4. Preview table schemas and sample data
Before writing a transform or running an analysis, it helps to know what you're working with. KAI Assistant can pull table schemas and data previews on demand, so you can understand your data structure without running queries manually.
Try asking:
"What columns does the orders table have?"
"Show me a preview of the sessions table."
"What data types are used in the product_catalog table?"
When sample data previews are enabled, data is converted to synthetic data before any LLM processing — so you get a useful preview without exposing raw customer data.
5. Create charts and graphs from your data
Once you're looking at a table or transform, you can ask KAI Assistant to visualize it. Describe what you want to see, and KAI Assistant generates a chart inline — directly inside the chat, without switching to another tool or exporting data.
Try asking:
"Create a bar chart of revenue by channel for the last 90 days."
"Show me a line graph of daily active users over the past month."
"Visualize order volume by region as a pie chart."
This is useful for getting a quick read on the data before building a full dashboard, or for sharing a snapshot with a stakeholder without any extra steps.
6. Debug pipeline issues using logs
Tracking down a failed pipeline run used to mean scrolling through raw logs and cross-referencing manually. KAI lets you search logs in plain English and get straight to the issue.
Try asking:
"Why did my daily revenue transform fail last night?"
"Show me any errors from the last 24 hours."
"What's the error message for the sessions pipeline?"
KAI Assistant surfaces the relevant log entries and — where possible — suggests what caused the issue and how to fix it.
A few things worth knowing
Your data stays private. KAI Assistant only works with your prompts, warehouse metadata, and (where enabled) synthetic previews of sample data. Raw customer data is never sent to an LLM. Kleene.ai does not use any customer data to train models.
KAI Assistant is context-aware. It knows where you are in the platform, so you don't need to over-explain your questions. If you're working inside a transform, it already knows the relevant context.
It's Phase 1. What's available now focuses on SQL generation, pipeline navigation, log debugging, and documentation. More capabilities are coming — including deeper pipeline-level intelligence and proactive anomaly detection.
Where to go from here
The best way to get value from KAI Assistant is to start using it on a real task — a transform you need to write, a pipeline that failed, a question you'd normally Google.
%201.svg)
