How It Works
When you ask a question, the AI Analyst:- Classifies your question — Determines whether you need data analysis, a definition, or a diagnostic check
- Identifies relevant tables — Finds the right data sources for your question
- Generates SQL — Writes a query against your BigQuery warehouse
- Executes and visualizes — Runs the query and creates a chart if appropriate
The AI Analyst queries your actual data in BigQuery. Results reflect your real business metrics, not sample data.
What You Can Ask
The AI Analyst handles three types of questions:Data Questions
Questions that require querying your warehouse:- “What was our revenue last week?”
- “Show me top 10 products by units sold this month”
- “How many new customers did we acquire in January?”
- “What’s our average order value by channel?”
Definitions & Schema
Questions about metrics, tables, data structure, and how SourceMedium works:- “What is LTV?”
- “How is conversion rate calculated?”
- “Which table has order data?”
- “What columns are in the customers table?”
- “How does attribution work in SourceMedium?”
- “What’s the difference between first-click and last-click attribution?”
Diagnostics
Health checks for your data and tracking:- “How is my data health?” — Check table freshness and availability
- “How is my attribution health?” — Check UTM tracking coverage
Data Health
Understand whether your data is fresh and ready for analysis.
Attribution Health
Diagnose tracking gaps and improve attribution coverage.
Analytical Domains
The AI Analyst can query data across these domains:| Domain | What You Can Analyze |
|---|---|
| Orders & Revenue | Revenue, AOV, order counts, discounts, shipping |
| Customers | New vs. repeat, cohort analysis, customer-level metrics |
| Marketing | Ad spend, ROAS, CPO, channel performance |
| Email & SMS | Campaign performance, engagement metrics |
| Web Analytics | Funnel events, page views, conversion paths |
| Cohort LTV | Lifetime value by acquisition cohort |
| Attribution | Multi-touch attribution, channel contribution |
Analysis Workflows
The AI Analyst uses different workflows depending on your question type:| Question Type | Workflow |
|---|---|
| Specific metrics | Standard Analysis |
| Open-ended/strategic | Deep Analysis |
| Definitions & schema | Knowledge Retrieval |
Learn about Workflows
Understand how different question types are routed and processed.
Deep Analysis
For complex, open-ended questions — like “How can we improve our marketing performance?” or “What trends should we be aware of?” — the AI Analyst automatically engages Deep Analysis. In this mode, the AI breaks your question into multiple analytical perspectives, runs parallel Standard Analyses, and synthesizes the findings into a comprehensive answer.Deep Analysis Details
Learn how Deep Analysis works, when it activates, and how to get the best results.
Feedback
Every response includes a feedback button. Use it to:- Rate whether the answer was helpful
- Report issues or inaccuracies
- Suggest improvements
Getting Started
1
Install the app
Visit your tenant URL (
https://{your-tenant}.sourcemedium.com/slack/install) and click “Add to Slack” to install the AI Analyst in your workspace. See Setup & Installation for details.2
Invite to a channel
Add the bot to a channel where you want to ask questions, or send it a direct message.
3
Ask a question
Type your question in natural language. Be specific about time ranges and metrics when possible.
4
Review results
The AI Analyst returns a response with data, a chart (if applicable), and the SQL query used.
What It Can’t Do
The AI Analyst is designed for analytics questions against your SourceMedium data. It cannot:- Access systems outside your BigQuery warehouse
- Modify data or run write operations
- Answer questions unrelated to your e-commerce analytics
- Guarantee real-time data (see Data Health for freshness)
- Access custom tables or customized logic built on top of SourceMedium data
Data Access & Security
Data Access Scope
Currently, the AI Analyst has access only to SourceMedium out-of-the-box transformed tables. If you have built custom tables or added customized logic on top of our data within your BigQuery instance, the agent will not have visibility into those for now.Support for custom tables and user-defined schemas is planned for a future update.
Access Control
The AI Analyst operates at the tenant level. This means:- Full Tenant Visibility: Anyone within your Slack workspace who has access to the bot can query any data available in the SourceMedium tables for your tenant.
- Multi-Store Organizations: If your tenant includes multiple stores, the bot can query data across all of them.
Next Steps
Example Questions
See examples of questions across different domains.
Understanding Results
Learn how to read responses, charts, and SQL output.