Skip to main content

Getting Started

Once the Slack integration is installed, mention @Vigilos in any channel where the bot has been added, followed by your question:
@Vigilos What were the top 10 products by revenue last month?
The bot will process your question and reply in a thread with the results.

How It Works

When you mention @Vigilos with a question, the bot follows a multi-step process to deliver an answer:
1

Identity resolution

The bot matches your Slack email address to your Vigilos account. This determines which organization, connections, and semantic models you have access to.
2

Intent classification

The bot classifies your message as either a data question (requiring database queries) or a conversational message (greetings, follow-ups, clarifications). Conversational messages get a direct response without running the data pipeline.
3

Model selection

If your organization has multiple semantic models, the bot presents a selection UI so you can choose which model to query against. If only one model exists, it is selected automatically.
4

Query execution

The AI agent analyzes your question against the semantic model, generates an optimized query, executes it against your database, and analyzes the results.
5

Formatted response

The bot delivers a formatted response in the thread, including data tables, charts (when applicable), and an AI-generated explanation of the results.

Thread Conversations

The Vigilos bot maintains conversation context within a Slack thread. You can ask follow-up questions without repeating context:
@Vigilos What were total sales last quarter?
Then reply in the thread:
@Vigilos Break that down by region
@Vigilos Now show me the month-over-month trend
The bot remembers previous questions and results within the thread, so follow-up questions build on the existing context.

Progress Reporting

For queries that take more than a few seconds, the bot provides progress updates in the thread so you know the analysis is underway. You will see status indicators as the bot works through schema fetching, query generation, and execution.

Supported Queries

The Slack bot supports the same types of questions as the web-based Ask AI interface:
Query TypeExample
Aggregations”What is the total revenue this month?”
Rankings”Top 5 customers by order count”
Trends”Show me daily signups over the past 30 days”
Comparisons”Compare revenue between Q1 and Q2”
Filters”How many orders from Germany had a value over $100?”
Breakdowns”Revenue by product category and region”

Tips for Better Results

Instead of “recently”, say “last 7 days” or “in January 2025”. Specific time ranges produce more accurate queries.
The bot works best when you use the entity and column names defined in your semantic model. For example, if your model has a “Customer Lifetime Value” measure, use that exact term.
Ask a broad question first, then use follow-up messages in the thread to drill down. This helps the bot understand exactly what you are looking for.
If you have multiple semantic models, make sure you select the one that covers the data you are asking about. A sales model will not be able to answer questions about engineering metrics.

Limitations

While the Slack bot handles most analytical questions, some scenarios work better in the Vigilos web app:
  • Complex multi-step analyses that require iterating on intermediate results
  • Visual builder configurations where you want fine-grained control over chart types and formatting
  • Saving and sharing insights - results from Slack are not automatically saved as insights
  • Dashboard creation - use the web app to build and organize dashboards
If the bot cannot answer a question in Slack, try the same question in the Ask AI interface on the web, which provides more interactive tools for complex analysis.