Tutorial: Building a Customer Churn Monitoring Agent

This tutorial walks you through building a practical business automation agent that monitors customer churn risk and proactively alerts your sales team to at-risk accounts.

What is "churn"? In business, customer churn refers to customers who stop doing business with you. A churn monitoring agent analyzes transaction patterns to identify customers showing signs of disengagement before they leave - enabling proactive retention efforts.

What You'll Build

A Customer Churn Monitoring Agent that:

Queries customer transaction data from your database
Groups and analyzes revenue trends per customer
Identifies customers with declining engagement (>20% decline)
Uses AI to generate personalized retention recommendations (formatted as HTML)
Emails a weekly HTML report to your sales team

Business Value: Early warning system enabling proactive retention efforts before customers churn.

Target Audience: Sales Operations, RevOps, Customer Success teams.

Prerequisites

Access to the Datafi platform
A data source with customer transaction data configured
Email service configured for notifications

Reference: Complete Agent Spec

This tutorial includes a fully working agent specification that you can:

Import directly - Load via the addAgent API and run immediately
Use as a template - Copy and modify for similar workflows
Compare your work - Verify your agent matches the expected structure

Important: The JSON specification in the Appendix is the authoritative reference. The conversational builder will generate a starting point (typically 60-70% complete), which you'll then refine using the visual editor or by editing the JSON directly.

Part 1: Understanding the Workflow

Workflow Overview

Start
  │
  ▼
[Query] ── Fetch customer transactions with monthly aggregation
  │
  ▼
[Array: Map] ── Add month index for trend analysis
  │
  ▼
[Regression] ── Statistical trend analysis (optional visualization)
  │
  ▼
[JSON: Query] ── Group by customer, calculate decline percentage
  │
  ▼
[Array: Filter] ── Filter customers with >20% decline
  │
  ▼
[JSON: Query] ── Count at-risk customers
  │
  ▼
[Condition] ── Any at-risk customers found?
  │
  ├── No → [End]
  │
  ▼ Yes
[LLM] ── Generate HTML-formatted retention recommendations
  │
  ▼
[Table Formatter] ── Format as HTML table (for email)
  │
  ▼
[Email] ── Send HTML report to sales team
  │
  ▼
End

Tools Used

Tool	Purpose
`query`	Execute PRQL queries against your data sources
`array`	Map and filter operations on arrays
`json`	JQ expressions for grouping, counting, transforming
`regression`	Statistical trend analysis
`llm`	AI-powered recommendation generation
`markdown_table_formatter`	Format data as HTML or markdown tables
`email`	Send HTML/text notifications

Node Types

Start: Entry point of the workflow
Action: Executes a tool (query, email, etc.)
Condition: Branches based on logic
End: Exit point of the workflow

Part 2: Using the Conversational Builder

The agent builder uses a conversational AI interface to help you create an initial workflow scaffold through natural language.

Setting Expectations: The AI builder will generate a solid starting point - typically 60-70% of what you need. You'll then refine the workflow using the visual editor (Part 3) to add details like HTML output formatting, proper variable references, and edge cases.

Step 1: Start a New Session

Navigate to the Agent Builder and start a new session with your initial description:

"Create an agent that monitors customer churn risk and alerts the sales team
when customers show declining engagement"

The AI will acknowledge your request and ask clarifying questions.

Step 2: Clarify Your Data Source

AI asks: "What data source contains your customer transaction history?"

Your response:

"Sales transactions from our database. I need to join Sales.CustomerTransactions
with Sales.Customers to get customer names. Key columns are CustomerID,
TransactionDate, TransactionAmount, and CustomerName."

Step 3: Define At-Risk Criteria

AI asks: "How should I identify at-risk customers?"

Your response:

"Calculate each customer's monthly revenue totals, then identify customers
whose revenue has declined by more than 20% from their first month to their
most recent month."

Step 4: Specify Output Format

AI asks: "What should happen when at-risk customers are found?"

Your response:

"Generate AI recommendations for each customer (formatted as HTML for email),
create an HTML table summarizing the at-risk customers, and email the report
to [email protected]"

Step 5: Set the Schedule

AI asks: "How often should this agent run?"

Your response:

"Weekly on Monday mornings at 8 AM Eastern time"

Step 6: Generate the Workflow

When the confidence indicator reaches the threshold (typically 80%+), click "Generate Workflow".

The system will create a graph-based workflow that you'll refine in the next step.

Part 3: Refining in the Visual Editor

After generation, the visual editor displays your workflow as a connected graph. This is where the real work happens - refining the AI's output into a production-ready agent.

1. Fix Data Transformations

The AI might generate placeholder transformations. Update them with actual JQ expressions:

Group by customer and calculate decline:

{
  "tool": "json",
  "params": {
    "operation": "query",
    "data": "${with_month_index}",
    "expression": "group_by(.customer_id) | map({customer_id: .[0].customer_id, customer_name: .[0].customer_name, first_month: .[0].monthly_total, last_month: .[-1].monthly_total, peak_month: (map(.monthly_total) | max), decline_percentage: (((.[0].monthly_total - .[-1].monthly_total) / (if .[0].monthly_total == 0 then 1 else .[0].monthly_total end)) * 100)})"
  }
}

2. Add Missing Nodes

You may need to add nodes the AI missed:

Count node (for displaying count in email):

{
  "tool": "json",
  "params": {
    "operation": "query",
    "data": "${at_risk}",
    "expression": "length"
  },
  "outputVar": "at_risk_count"
}

3. Configure HTML Output

For email-compatible output, set the table formatter to HTML:

{
  "tool": "markdown_table_formatter",
  "params": {
    "data": "${at_risk}",
    "columns": ["customer_name", "first_month", "last_month", "decline_percentage"],
    "columnLabels": ["Customer", "First Month Revenue", "Last Month Revenue", "Decline %"],
    "valueFormatting": {
      "first_month": {"prefix": "$"},
      "last_month": {"prefix": "$"},
      "decline_percentage": {"suffix": "%"}
    },
    "outputFormat": "html"
  }
}

4. Configure LLM for HTML Output

Tell the LLM to output HTML (not markdown) for proper email rendering:

{
  "tool": "llm",
  "params": {
    "messages": [
      {
        "role": "system",
        "content": "You are a customer success expert. Generate specific, actionable retention recommendations based on customer data. Format your response as HTML (use <h4>, <p>, <ul>, <li>, <strong> tags). Do NOT use markdown syntax."
      },
      {
        "role": "user",
        "content": "Analyze these at-risk customers and provide 2-3 retention recommendations for each:\n\n${at_risk}\n\nFor each customer, consider their decline percentage and transaction history. Output as HTML."
      }
    ]
  }
}

5. Build the Email Template

Construct the HTML email body with variable substitution:

{
  "tool": "email",
  "params": {
    "to": "${email_recipients}",
    "subject": "Weekly At-Risk Customer Alert - ${at_risk_count} customers need attention",
    "html": "<h2>Weekly At-Risk Customer Alert</h2>\n<p><strong>Report Date:</strong> ${report_date}</p>\n<p><strong>At-Risk Customers Found:</strong> ${at_risk_count}</p>\n\n<h3>At-Risk Customer Summary</h3>\n${report_table}\n\n<h3>Retention Recommendations</h3>\n<div>${recommendations.response}</div>\n\n<hr>\n<p><em>Generated by Customer Churn Monitoring Agent</em></p>"
  }
}

Key insight: Use ${recommendations.response} to extract just the text from the LLM response object, not the full JSON.

Testing Your Workflow

Click "Run Now" to execute manually
Watch real-time progress via the execution panel
Review each node's output - check for errors
Verify the email renders correctly (HTML table, formatted recommendations)

Part 4: Key Technical Details

PRQL Query Structure

The query node fetches transactions with monthly aggregation:

from t = Sales.CustomerTransactions
join c = Sales.Customers (t.CustomerID == c.CustomerID)
select {
  transaction_date = t.TransactionDate,
  customer_id = t.CustomerID,
  customer_name = c.CustomerName,
  amount = t.TransactionAmount,
  year = s"YEAR(TransactionDate)",
  month = s"MONTH(TransactionDate)"
}
filter transaction_date >= '${start_date}'
group {customer_id, customer_name, year, month} (
  aggregate {
    monthly_total = sum amount,
    transaction_count = count this
  }
)
sort {customer_id, year, month}

Variable Substitution Patterns

Pattern	Use Case
`${variable_name}`	Simple variable substitution
`${object.property}`	Access nested properties (e.g., `${recommendations.response}`)
`${at_risk_count}`	Use computed values (requires separate count node)

Note: ${array.length} works in Condition nodes but NOT in string templates. Create a separate count node for email subjects/body.

Filter Expression Syntax

The array tool uses JQ-style expressions:

{
  "operation": "filter",
  "data": "${customer_trends}",
  "expression": ".decline_percentage > 20"
}

Percentage Calculation

Store percentages as actual numbers (100 = 100%, not 1.0):

decline_percentage: (((first - last) / first) * 100)

Then format with suffix in the table:

"valueFormatting": {
  "decline_percentage": {"suffix": "%"}
}

Part 5: Running and Monitoring

Manual Execution

Navigate to your agent in the dashboard
Click "Run Now"
Watch execution progress in real-time
View node-by-node output in the execution log

Scheduled Execution

Configure the schedule in the agent's lifecycle settings:

{
  "trigger": {
    "Schedule": {
      "cron": "0 8 * * MON",
      "timezone": "America/New_York"
    }
  }
}

Cron expression breakdown:

0 - At minute 0
8 - At hour 8 (8 AM)
* - Every day of month
* - Every month
MON - On Mondays

Monitoring Execution History

Track your agent's performance:

Run History: View past executions, success/failure status
Execution Logs: Detailed output from each node
Token Usage: Monitor LLM token consumption
Error Tracking: Identify and debug failures

Part 6: Next Steps

Customize for Your Data

Replace the placeholder table names with your actual schema:

-- Template (in sample spec)
from t = Sales.CustomerTransactions
join c = Sales.Customers (t.CustomerID == c.CustomerID)

-- Your data (customize)
from t = YourSchema.Orders
join c = YourSchema.Customers (t.customer_id == c.id)

Add Human Approval

Insert a HumanInput node before the email step to review the report:

{
  "type": "HumanInput",
  "data": {
    "label": "Review At-Risk Report",
    "prompt": "Please review the at-risk customer report before sending to the sales team.",
    "options": ["Approve and Send", "Edit Report", "Cancel"]
  }
}

Expand Your Analysis

Add more sophisticated churn indicators:

Support ticket analysis: Flag customers with increased support requests
Login frequency: Track declining platform usage
NPS scores: Incorporate customer satisfaction data
Contract renewal dates: Prioritize customers approaching renewal

Build More Agents

Apply the same patterns to other business problems:

Lead Scoring Agent: Score and prioritize inbound leads
Revenue Forecasting Agent: Generate weekly revenue projections
Competitor Monitoring Agent: Track competitor pricing and features
Invoice Reconciliation Agent: Automate invoice matching and discrepancy detection

Summary

You've learned how to:

Understand the workflow structure - Query, transform, analyze, act
Use the conversational builder - Generate a 60-70% starting point
Refine in the visual editor - Add HTML formatting, fix transformations, handle edge cases
Configure for production - Proper variable substitution, HTML output, email templates
Monitor and iterate - Test, debug, and improve

The key insight: The JSON specification is your source of truth. The conversational builder accelerates creation, but production-quality agents require hands-on refinement.

For more information, see:

Agent Builder - API reference
Workflow Builder - Real-time execution
AI Infrastructure Overview - System overview

Appendix: Complete Agent Specification

Below is the complete, production-ready agent specification. This is the gold standard - import it directly via the addAgent API to see a fully working agent, or use it as a reference when building your own.

Key Configuration Highlights

Workflow Variables (customize these for your environment):

{
  "start_date": "2024-01-01",
  "decline_threshold": 20,
  "email_recipients": "[email protected]",
  "report_date": "2024-03-20"
}

Critical Implementation Details:

Decline percentage stored as actual percentage (100, not 1.0)
HTML output format for email-compatible tables
LLM instructed to output HTML (not markdown)
Separate count node for ${at_risk_count} in email
Use ${recommendations.response} to extract LLM text from response object

Schedule Trigger:

{
  "trigger": {
    "Schedule": {
      "cron": "0 8 * * MON",
      "timezone": "America/New_York"
    }
  }
}

Workflow Node Summary

Node ID	Tool	Purpose
`query_transactions`	query	Fetch and aggregate transaction data
`add_month_index`	array (map)	Add month index for sorting
`analyze_trends`	regression	Statistical trend visualization
`calculate_trends`	json (query)	Group by customer, calculate decline %
`filter_at_risk`	array (filter)	Filter customers >20% decline
`count_at_risk`	json (query)	Count for email subject/body
`check_at_risk`	condition	Branch if any at-risk found
`generate_recommendations`	llm	AI recommendations (HTML format)
`format_report`	markdown_table_formatter	HTML table for email
`send_email`	email	Send HTML report

Full Agent Specification (JSON)

{
  "identity": {
    "name": "customer-churn-monitor",
    "version": "1.0.0",
    "description": "Analyzes customer transaction patterns to detect churn risk (customers likely to stop doing business). Identifies declining revenue trends, generates AI-powered retention recommendations, and sends weekly alerts to the sales team.",
    "author": "Datafi",
    "tags": ["sales", "customer-success", "churn-prevention", "analytics"],
    "goals": [
      "Identify customers with declining revenue trends",
      "Generate actionable retention recommendations",
      "Deliver weekly alerts to sales team"
    ],
    "successCriteria": [
      {
        "metric": "at_risk_detection_rate",
        "target": 0.9,
        "measurement": "detected_churned / actual_churned"
      },
      {
        "metric": "report_delivery",
        "target": "100%",
        "measurement": "reports_sent / reports_scheduled"
      }
    ]
  },
  "capabilities": {
    "tools": [
      { "name": "query", "permissions": { "allowedOperations": ["Read"] } },
      { "name": "array", "permissions": { "allowedOperations": ["Read"] } },
      { "name": "json", "permissions": { "allowedOperations": ["Read"] } },
      { "name": "regression", "permissions": { "allowedOperations": ["Read"] } },
      { "name": "llm", "permissions": { "allowedOperations": ["Read"] } },
      { "name": "markdown_table_formatter", "permissions": { "allowedOperations": ["Read"] } },
      { "name": "email", "permissions": { "allowedOperations": ["Write"] } }
    ],
    "skills": ["trend_analysis", "churn_prediction", "report_generation"],
    "maxIterations": 25,
    "supportedDataSources": [],
    "outputFormats": []
  },
  "behavior": {
    "executionMode": "Sequential",
    "retryPolicy": {
      "maxAttempts": 3,
      "backoffStrategy": "Exponential",
      "initialDelaySeconds": 5,
      "maxDelaySeconds": 60,
      "retryOn": ["network_error", "timeout", "rate_limit"],
      "doNotRetry": ["validation_error", "permission_denied"]
    },
    "timeoutSeconds": 300,
    "concurrencyLimit": 1,
    "personality": {
      "tone": "professional",
      "verbosity": "concise",
      "jargonLevel": "business_appropriate",
      "interactionStyle": "proactive"
    },
    "reasoningStrategy": {
      "approach": "StepByStep",
      "explorationMode": "depth_first",
      "confidenceThreshold": 0.75,
      "explainReasoning": true
    }
  },
  "guards": {
    "constraints": [
      "Never modify production data",
      "PII must be masked in email outputs",
      "Only read from approved data sources"
    ],
    "validationRules": [],
    "outputFilters": [],
    "resourceLimits": {
      "maxTokensPerRequest": 10000,
      "maxApiCallsPerRun": 100,
      "dailyTokenBudget": 50000
    },
    "rateLimits": {},
    "sensitiveDataHandling": {
      "classificationLevels": ["Internal"],
      "encryptionRequired": [],
      "auditAccess": true
    },
    "approvalRequirements": [],
    "allowedDatabases": [],
    "blockedDatabases": [],
    "preventPiiExposure": false,
    "preventSqlInjection": true
  },
  "testing": {
    "testScenarios": [],
    "mockData": {},
    "benchmarks": [],
    "evaluationMetrics": [],
    "regressionTests": [],
    "mockEnvironments": [],
    "benchmarkDatasets": []
  },
  "lifecycle": {
    "trigger": {
      "Schedule": {
        "cron": "0 8 * * MON",
        "timezone": "America/New_York"
      }
    },
    "dependencies": [],
    "versioningStrategy": "",
    "deprecationPolicy": null,
    "updateMechanism": null
  },
  "integration": {
    "inputSchema": {
      "type": "object",
      "properties": {
        "lookback_months": {
          "type": "integer",
          "default": 6,
          "description": "Number of months to analyze"
        },
        "decline_threshold": {
          "type": "number",
          "default": 20,
          "description": "Minimum decline percentage to flag (e.g., 20 = 20%)"
        },
        "email_recipients": {
          "type": "string",
          "default": "[email protected]",
          "description": "Email address for alerts"
        }
      }
    },
    "outputSchema": {
      "type": "object",
      "properties": {
        "at_risk_count": { "type": "integer" },
        "report_sent": { "type": "boolean" }
      }
    },
    "webhooks": [],
    "eventEmitters": [],
    "apiEndpoints": []
  },
  "monitoring": {
    "metrics": [
      { "metric": "customers_analyzed", "type": "Counter" },
      { "metric": "at_risk_detected", "type": "Counter" },
      { "metric": "execution_duration_seconds", "type": "Histogram" }
    ],
    "alerts": [],
    "logging": { "level": "Info" },
    "tracing": {}
  },
  "workflow": {
    "nodes": [
      {
        "id": "start_0",
        "type": "agentFlow",
        "position": { "x": 100, "y": 100 },
        "data": {
          "id": "start_0",
          "type": "Start",
          "label": "Start",
          "name": "Start",
          "outputAnchors": [{ "id": "start_0-out", "name": "output" }]
        }
      },
      {
        "id": "query_transactions",
        "type": "agentFlow",
        "position": { "x": 100, "y": 200 },
        "data": {
          "id": "query_transactions",
          "type": "Action",
          "label": "Query Customer Transactions",
          "name": "Query Customer Transactions",
          "inputs": {
            "tool": "query",
            "params": {
              "query": "from t = Sales.CustomerTransactions\njoin c = Sales.Customers (t.CustomerID == c.CustomerID)\nselect {\n  transaction_date = t.TransactionDate,\n  customer_id = t.CustomerID,\n  customer_name = c.CustomerName,\n  amount = t.TransactionAmount,\n  year = s\"YEAR(TransactionDate)\",\n  month = s\"MONTH(TransactionDate)\"\n}\nfilter transaction_date >= '${start_date}'\ngroup {customer_id, customer_name, year, month} (\n  aggregate {\n    monthly_total = sum amount,\n    transaction_count = count this\n  }\n)\nsort {customer_id, year, month}"
            },
            "outputVar": "transactions"
          },
          "outputAnchors": [{ "id": "query_transactions-out", "name": "output" }]
        }
      },
      {
        "id": "add_month_index",
        "type": "agentFlow",
        "position": { "x": 100, "y": 300 },
        "data": {
          "id": "add_month_index",
          "type": "Action",
          "label": "Add Month Index",
          "name": "Add Month Index",
          "inputs": {
            "tool": "array",
            "params": {
              "operation": "map",
              "data": "${transactions}",
              "expression": ". + {month_index: (.year * 12 + .month)}"
            },
            "outputVar": "with_month_index"
          },
          "outputAnchors": [{ "id": "add_month_index-out", "name": "output" }]
        }
      },
      {
        "id": "analyze_trends",
        "type": "agentFlow",
        "position": { "x": 100, "y": 400 },
        "data": {
          "id": "analyze_trends",
          "type": "Action",
          "label": "Analyze Revenue Trends",
          "name": "Analyze Revenue Trends",
          "inputs": {
            "tool": "regression",
            "params": {
              "data": "${with_month_index}",
              "groupBy": "customer_id",
              "xColumn": "month_index",
              "yColumn": "monthly_total",
              "confidenceLevel": 0.95
            },
            "outputVar": "regression_result"
          },
          "outputAnchors": [{ "id": "analyze_trends-out", "name": "output" }]
        }
      },
      {
        "id": "calculate_trends",
        "type": "agentFlow",
        "position": { "x": 100, "y": 500 },
        "data": {
          "id": "calculate_trends",
          "type": "Action",
          "label": "Calculate Customer Trends",
          "name": "Calculate Customer Trends",
          "inputs": {
            "tool": "json",
            "params": {
              "operation": "query",
              "data": "${with_month_index}",
              "expression": "group_by(.customer_id) | map({customer_id: .[0].customer_id, customer_name: .[0].customer_name, first_month: .[0].monthly_total, last_month: .[-1].monthly_total, peak_month: (map(.monthly_total) | max), decline_percentage: (((.[0].monthly_total - .[-1].monthly_total) / (if .[0].monthly_total == 0 then 1 else .[0].monthly_total end)) * 100)})"
            },
            "outputVar": "customer_trends"
          },
          "outputAnchors": [{ "id": "calculate_trends-out", "name": "output" }]
        }
      },
      {
        "id": "filter_at_risk",
        "type": "agentFlow",
        "position": { "x": 100, "y": 700 },
        "data": {
          "id": "filter_at_risk",
          "type": "Action",
          "label": "Filter At-Risk Customers",
          "name": "Filter At-Risk Customers",
          "inputs": {
            "tool": "array",
            "params": {
              "operation": "filter",
              "data": "${customer_trends}",
              "expression": ".decline_percentage > 20"
            },
            "outputVar": "at_risk"
          },
          "outputAnchors": [{ "id": "filter_at_risk-out", "name": "output" }]
        }
      },
      {
        "id": "count_at_risk",
        "type": "agentFlow",
        "position": { "x": 100, "y": 750 },
        "data": {
          "id": "count_at_risk",
          "type": "Action",
          "label": "Count At-Risk Customers",
          "name": "Count At-Risk Customers",
          "inputs": {
            "tool": "json",
            "params": {
              "operation": "query",
              "data": "${at_risk}",
              "expression": "length"
            },
            "outputVar": "at_risk_count"
          },
          "outputAnchors": [{ "id": "count_at_risk-out", "name": "output" }]
        }
      },
      {
        "id": "check_at_risk",
        "type": "agentFlow",
        "position": { "x": 100, "y": 800 },
        "data": {
          "id": "check_at_risk",
          "type": "Condition",
          "label": "Any At-Risk Customers?",
          "name": "Any At-Risk Customers?",
          "inputs": {
            "condition": "${at_risk_count} > 0",
            "params": {}
          },
          "outputAnchors": [
            { "id": "check_at_risk-true", "name": 0, "label": "True" },
            { "id": "check_at_risk-false", "name": 1, "label": "False" }
          ]
        }
      },
      {
        "id": "generate_recommendations",
        "type": "agentFlow",
        "position": { "x": 100, "y": 900 },
        "data": {
          "id": "generate_recommendations",
          "type": "Action",
          "label": "Generate Retention Recommendations",
          "name": "Generate Retention Recommendations",
          "inputs": {
            "tool": "llm",
            "params": {
              "messages": [
                {
                  "role": "system",
                  "content": "You are a customer success expert. Generate specific, actionable retention recommendations based on customer data. Format your response as HTML (use <h4>, <p>, <ul>, <li>, <strong> tags). Do NOT use markdown syntax."
                },
                {
                  "role": "user",
                  "content": "Analyze these at-risk customers and provide 2-3 retention recommendations for each:\n\n${at_risk}\n\nFor each customer, consider their decline percentage and transaction history. Output as HTML."
                }
              ]
            },
            "outputVar": "recommendations"
          },
          "outputAnchors": [{ "id": "generate_recommendations-out", "name": "output" }]
        }
      },
      {
        "id": "format_report",
        "type": "agentFlow",
        "position": { "x": 100, "y": 1000 },
        "data": {
          "id": "format_report",
          "type": "Action",
          "label": "Format Report",
          "name": "Format Report",
          "inputs": {
            "tool": "markdown_table_formatter",
            "params": {
              "data": "${at_risk}",
              "columns": ["customer_name", "first_month", "last_month", "decline_percentage"],
              "columnLabels": ["Customer", "First Month Revenue", "Last Month Revenue", "Decline %"],
              "sortBy": ["decline_percentage"],
              "valueFormatting": {
                "first_month": { "prefix": "$" },
                "last_month": { "prefix": "$" },
                "decline_percentage": { "suffix": "%" }
              },
              "outputFormat": "html"
            },
            "outputVar": "report_table"
          },
          "outputAnchors": [{ "id": "format_report-out", "name": "output" }]
        }
      },
      {
        "id": "send_email",
        "type": "agentFlow",
        "position": { "x": 100, "y": 1100 },
        "data": {
          "id": "send_email",
          "type": "Action",
          "label": "Send Alert Email",
          "name": "Send Alert Email",
          "inputs": {
            "tool": "email",
            "params": {
              "to": "${email_recipients}",
              "subject": "Weekly At-Risk Customer Alert - ${at_risk_count} customers need attention",
              "html": "<h2>Weekly At-Risk Customer Alert</h2>\n<p><strong>Report Date:</strong> ${report_date}</p>\n<p><strong>At-Risk Customers Found:</strong> ${at_risk_count}</p>\n\n<h3>At-Risk Customer Summary</h3>\n${report_table}\n\n<h3>Retention Recommendations</h3>\n<div>${recommendations.response}</div>\n\n<hr>\n<p><em>Generated by Customer Churn Monitoring Agent</em></p>"
            }
          },
          "outputAnchors": [{ "id": "send_email-out", "name": "output" }]
        }
      },
      {
        "id": "end_success",
        "type": "agentFlow",
        "position": { "x": 175, "y": 1400 },
        "data": {
          "id": "end_success",
          "type": "End",
          "label": "End (Report Sent)",
          "name": "End",
          "outputAnchors": []
        }
      },
      {
        "id": "end_no_risk",
        "type": "agentFlow",
        "position": { "x": 300, "y": 900 },
        "data": {
          "id": "end_no_risk",
          "type": "End",
          "label": "End (No At-Risk)",
          "name": "End",
          "outputAnchors": []
        }
      }
    ],
    "edges": [
      { "id": "e1", "source": "start_0", "target": "query_transactions", "type": "workflow", "data": {} },
      { "id": "e2", "source": "query_transactions", "target": "add_month_index", "sourceHandle": "query_transactions-out", "type": "workflow", "data": {} },
      { "id": "e3", "source": "add_month_index", "target": "analyze_trends", "sourceHandle": "add_month_index-out", "type": "workflow", "data": {} },
      { "id": "e3b", "source": "analyze_trends", "target": "calculate_trends", "sourceHandle": "analyze_trends-out", "type": "workflow", "data": {} },
      { "id": "e4", "source": "calculate_trends", "target": "filter_at_risk", "sourceHandle": "calculate_trends-out", "type": "workflow", "data": {} },
      { "id": "e5", "source": "filter_at_risk", "target": "count_at_risk", "sourceHandle": "filter_at_risk-out", "type": "workflow", "data": {} },
      { "id": "e5b", "source": "count_at_risk", "target": "check_at_risk", "sourceHandle": "count_at_risk-out", "type": "workflow", "data": {} },
      { "id": "e6", "source": "check_at_risk", "target": "generate_recommendations", "sourceHandle": "check_at_risk-true", "type": "workflow", "data": {} },
      { "id": "e7", "source": "check_at_risk", "target": "end_no_risk", "sourceHandle": "check_at_risk-false", "type": "workflow", "data": {} },
      { "id": "e8", "source": "generate_recommendations", "target": "format_report", "sourceHandle": "generate_recommendations-out", "type": "workflow", "data": {} },
      { "id": "e9", "source": "format_report", "target": "send_email", "sourceHandle": "format_report-out", "type": "workflow", "data": {} },
      { "id": "e10", "source": "send_email", "target": "end_success", "sourceHandle": "send_email-out", "type": "workflow", "data": {} }
    ],
    "errorHandling": "FailFast",
    "variables": {
      "start_date": "2024-01-01",
      "decline_threshold": 20,
      "report_date": "2024-03-20",
      "email_recipients": "[email protected]"
    }
  }
}

What You'll Build​

Prerequisites​

Reference: Complete Agent Spec​

Part 1: Understanding the Workflow​

Workflow Overview​

Tools Used​

Node Types​

Part 2: Using the Conversational Builder​

Step 1: Start a New Session​

Step 2: Clarify Your Data Source​

Step 3: Define At-Risk Criteria​

Step 4: Specify Output Format​

Step 5: Set the Schedule​

Step 6: Generate the Workflow​

Part 3: Refining in the Visual Editor​

Common Refinements You'll Make​

1. Fix Data Transformations​

2. Add Missing Nodes​

3. Configure HTML Output​

4. Configure LLM for HTML Output​

5. Build the Email Template​

Testing Your Workflow​

Part 4: Key Technical Details​

PRQL Query Structure​

Variable Substitution Patterns​

Filter Expression Syntax​

Percentage Calculation​

Part 5: Running and Monitoring​

Manual Execution​

Scheduled Execution​

Monitoring Execution History​

Part 6: Next Steps​

Customize for Your Data​

Add Human Approval​

Expand Your Analysis​

Build More Agents​

Summary​

Appendix: Complete Agent Specification​

Key Configuration Highlights​

Workflow Node Summary​

Full Agent Specification (JSON)​

What You'll Build

Prerequisites

Reference: Complete Agent Spec

Part 1: Understanding the Workflow

Workflow Overview

Tools Used

Node Types

Part 2: Using the Conversational Builder

Step 1: Start a New Session

Step 2: Clarify Your Data Source

Step 3: Define At-Risk Criteria

Step 4: Specify Output Format

Step 5: Set the Schedule

Step 6: Generate the Workflow

Part 3: Refining in the Visual Editor

Common Refinements You'll Make

1. Fix Data Transformations

2. Add Missing Nodes

3. Configure HTML Output

4. Configure LLM for HTML Output

5. Build the Email Template

Testing Your Workflow

Part 4: Key Technical Details

PRQL Query Structure

Variable Substitution Patterns

Filter Expression Syntax

Percentage Calculation

Part 5: Running and Monitoring

Manual Execution

Scheduled Execution

Monitoring Execution History

Part 6: Next Steps

Customize for Your Data

Add Human Approval

Expand Your Analysis

Build More Agents

Summary

Appendix: Complete Agent Specification

Key Configuration Highlights

Workflow Node Summary

Full Agent Specification (JSON)