Home / Features / AI Agents

AI Agents

Specialized AI assistants configured with business context, security rules, and domain expertise.

🤖 What Are AI Agents?

AI Agents are specialized versions of the query engine that:

Understand context: Know your business metrics, terminology, and data relationships
Enforce security: Only access authorized tables and columns
Optimize queries: Generate efficient SQL based on your schema
Provide insights: Analyze trends and suggest actionable recommendations

🎯 Agent Types

📊 General Data Analyst

Best for: Ad-hoc analysis, exploration, general queries

Capabilities:
- Cross-functional analysis
- General statistical queries
- Data exploration
- Multi-table joins

Example Questions:
- "Show me top customers by revenue"
- "What's the trend in user signups?"
- "Compare metrics across regions"

📈 Sales Performance Analyst

Best for: Revenue tracking, sales pipeline, deal analysis

Business Context:
- Fiscal year, quota periods
- Product lines and pricing
- Sales territories
- Commission structure

Example Questions:
- "Show me Q4 pipeline by stage"
- "Which reps are below quota?"
- "Average deal size by product"

🎯 Marketing Campaign Analyst

Best for: Campaign performance, attribution, ROI

Business Context:
- Marketing channels
- Campaign lifecycle
- Attribution models
- Customer journey stages

Example Questions:
- "ROI by campaign channel"
- "Cost per acquisition trends"
- "Email conversion rates"

💰 Financial Data Analyst

Best for: Financial reporting, budget analysis, forecasting

Business Context:
- Fiscal periods
- Cost centers
- Budget categories
- Revenue recognition rules

Example Questions:
- "Budget vs actual by department"
- "Cash flow projections"
- "Cost breakdown analysis"

⚙️ Operations Analyst

Best for: Process efficiency, resource utilization, KPIs

Business Context:
- Process workflows
- SLAs and targets
- Resource allocation
- Efficiency metrics

Example Questions:
- "Average fulfillment time"
- "Resource utilization rates"
- "Process bottlenecks"

📱 Product Analytics Specialist

Best for: User behavior, feature adoption, engagement

Business Context:
- User actions and events
- Feature flags
- Product roadmap
- Engagement metrics

Example Questions:
- "Feature adoption by cohort"
- "User retention rates"
- "Most used features"

🔧 Agent Configuration

System Prompt

The system prompt gives agents business intelligence:

Example - Sales Agent:

You are a sales performance analyst for Acme Corp.

Key Metrics:
- Revenue, ARR, MRR
- Deal size, win rate, sales cycle
- Pipeline coverage, velocity

Business Rules:
- Fiscal year: January 1
- Quota period: Quarterly
- Product lines: Enterprise ($50K+), Pro ($10K+), Starter ($1K+)
- Regions: Americas, EMEA, APAC

When analyzing:
1. Include time comparisons (MoM, YoY, QoQ)
2. Segment by region and product
3. Highlight trends and anomalies
4. Provide actionable insights

Data Source Assignment

Each agent connects to ONE data source:

Production database
Analytics warehouse
Specific application DB

Table Access Control

Select which tables the agent can query:

Example - Sales Agent

Allowed Tables:

✅ customers
✅ orders
✅ products
✅ sales_reps
❌ internal_admin
❌ employee_salaries

Column Permissions

Hide sensitive columns even in allowed tables:

customers table

Visible: customer_id, name, company, industry, region
Hidden: ssn, credit_card, internal_notes

👥 User Assignment

Control which users can use which agents:

Assignment Strategies

By Department

Sales team → Sales Agent
Marketing team → Marketing Agent

By Role

Analysts → All agents
Managers → Department-specific

By Data Access

Match agent permissions to user data access rights

Multi-Agent

Power users get access to multiple specialized agents

🎓 Best Practices

✅ Do's

Specific context: Include business metrics, terminology, and rules
Clear scope: Define what the agent should focus on
Least privilege: Only grant necessary table/column access
Actionable insights: Prompt agents to provide recommendations
Regular updates: Keep prompts current with business changes

❌ Don'ts

Generic prompts: "You are a helpful assistant" is too vague
Over-permissive: Don't grant access to all tables
Ignore security: Always hide PII and sensitive data
One-size-fits-all: Create specialized agents for different use cases

📊 Agent Performance

Specialized agents generate better queries because they:

🎯 Understand Context

Know your metrics and terminology

🔍 Optimize SQL

Use appropriate indexes and joins

📈 Suggest Insights

Proactively identify trends

🔒 Enforce Security

Automatic data access control

🔄 Agent Management

Creating Agents

Navigate to Agents
Click "+ New Agent"
Select agent type template
Configure system prompt
Assign data source
Set table/column permissions
Activate agent

Editing Agents

System prompt: Update business context
Permissions: Add/remove table access
Status: Activate/deactivate

Cloning Agents

Create variations for different teams:

Click agent to clone
Select "Clone"
Modify name and settings
Assign to different users

🚀 Advanced Use Cases

Multi-Agent Workflows

Use different agents for different stages of analysis:

Exploration Agent: Broad access for initial investigation
Specialized Agent: Deep dive into specific area
Reporting Agent: Standard metrics and dashboards

Agent Hierarchies

Executive Agent
├── Limited high-level metrics
└── Cross-functional view

Department Agents
├── Sales Agent (sales data)
├── Marketing Agent (campaign data)
└── Finance Agent (financial data)

Analyst Agents
└── Full access for data team

🚀 Next Steps

← Natural Language Queries Next: Query Templates →