Customer Onboarding Assistant

​

Overview

​

What the Agent Needs to Accomplish

• Track customer progress through onboarding milestones
• Monitor feature adoption and usage patterns
• Identify customers who are stuck or at risk
• Provide personalized onboarding guidance
• Recommend next steps based on customer profile
• Analyze onboarding effectiveness and time-to-value

​

How Pylar Helps

• Unified Onboarding View: Combining account data, usage metrics, and milestone tracking
• Real-time Progress: Querying current onboarding status across all customers
• Pattern Recognition: Identifying successful onboarding patterns
• Personalized Recommendations: Tailoring guidance based on customer profile
• Effectiveness Analysis: Measuring onboarding success metrics

​

Without Pylar vs With Pylar

​

Without Pylar

• ❌ Data scattered across product analytics, CRM, and onboarding tools
• ❌ Manual tracking of onboarding progress
• ❌ Difficult to identify at-risk customers
• ❌ No unified view of onboarding effectiveness
• ❌ Time-consuming to generate personalized recommendations
• ❌ Limited ability to correlate onboarding with outcomes

• 3-4 different system integrations
• Custom onboarding tracking logic
• Manual analysis of progress
• Complex correlation of data sources
• ~4-5 weeks development time

​

With Pylar

• ✅ Single endpoint for all onboarding data
• ✅ Real-time progress tracking
• ✅ Automated at-risk customer identification
• ✅ Unified onboarding effectiveness metrics
• ✅ Easy personalized recommendations
• ✅ Built-in analytics on onboarding patterns

• Connect 3-4 data sources (1 hour)
• Create onboarding views (2.5 hours)
• Build MCP tools with AI (1 hour)
• Connect to agent builder (15 minutes)
• Total: ~5 hours

​

Step-by-Step Implementation

​

Step 1: Connect Data Sources

1. Connect Product Analytics (Feature usage, user actions, engagement)
2. Connect CRM (Account data, customer profile, subscription tier)
3. Connect Onboarding System (Milestone tracking, completion status)
4. Connect Support System (Onboarding-related tickets)

​

Step 2: Create Onboarding Views

CREATE VIEW onboarding_progress AS
SELECT 
  c.customer_id,
  c.customer_name,
  c.account_created_date,
  DATEDIFF(CURRENT_DATE, c.account_created_date) as days_since_signup,
  c.subscription_tier,
  -- Milestone tracking
  o.milestone_1_completed,
  o.milestone_1_completion_date,
  o.milestone_2_completed,
  o.milestone_2_completion_date,
  o.milestone_3_completed,
  o.milestone_3_completion_date,
  o.milestone_4_completed,
  o.milestone_4_completion_date,
  -- Feature adoption
  u.features_activated_count,
  u.total_actions_count,
  u.active_days_count,
  u.last_activity_date,
  DATEDIFF(CURRENT_DATE, u.last_activity_date) as days_since_last_activity,
  -- Support tickets
  s.onboarding_tickets_count,
  s.last_ticket_date,
  -- Progress calculation
  (
    CASE WHEN o.milestone_1_completed THEN 25 ELSE 0 END +
    CASE WHEN o.milestone_2_completed THEN 25 ELSE 0 END +
    CASE WHEN o.milestone_3_completed THEN 25 ELSE 0 END +
    CASE WHEN o.milestone_4_completed THEN 25 ELSE 0 END
  ) as progress_percentage,
  -- Risk indicators
  CASE 
    WHEN DATEDIFF(CURRENT_DATE, c.account_created_date) > 14 
         AND o.milestone_1_completed = FALSE THEN 1
    ELSE 0
  END as stalled_flag,
  CASE 
    WHEN DATEDIFF(CURRENT_DATE, u.last_activity_date) > 7 THEN 1
    ELSE 0
  END as inactive_flag
FROM crm.customers c
LEFT JOIN onboarding.milestones o ON c.customer_id = o.customer_id
LEFT JOIN analytics.user_usage u ON c.customer_id = u.customer_id
LEFT JOIN support.onboarding_tickets s ON c.customer_id = s.customer_id;

CREATE VIEW onboarding_effectiveness AS
SELECT 
  o.*,
  -- Time to value metrics
  CASE 
    WHEN o.milestone_4_completed = TRUE 
    THEN DATEDIFF(o.milestone_4_completion_date, o.account_created_date)
    ELSE NULL
  END as days_to_complete,
  -- Success indicators
  CASE 
    WHEN o.progress_percentage = 100 
         AND o.days_since_last_activity <= 7 
         AND o.features_activated_count >= 3
    THEN 'Successful'
    WHEN o.progress_percentage < 50 
         AND o.days_since_last_activity > 7
    THEN 'At Risk'
    WHEN o.progress_percentage >= 50 
         AND o.progress_percentage < 100
    THEN 'In Progress'
    ELSE 'Needs Attention'
  END as onboarding_status,
  -- Recommended next steps
  CASE 
    WHEN o.milestone_1_completed = FALSE THEN 'Complete account setup'
    WHEN o.milestone_2_completed = FALSE THEN 'Import your first data'
    WHEN o.milestone_3_completed = FALSE THEN 'Create your first dashboard'
    WHEN o.milestone_4_completed = FALSE THEN 'Invite your team'
    ELSE 'Continue exploring advanced features'
  END as recommended_action
FROM onboarding_progress o;

CREATE VIEW at_risk_onboarding AS
SELECT 
  e.*
FROM onboarding_effectiveness e
WHERE e.onboarding_status IN ('At Risk', 'Needs Attention')
  AND e.days_since_signup <= 30
ORDER BY e.churn_risk_score DESC, e.days_since_signup DESC;

​

Step 3: Create MCP Tools with AI

• Prompt: “Create a tool to get onboarding progress for a customer by customer ID”
• AI generates: get_onboarding_progress(customer_id: string)

• Prompt: “Create a tool to list customers who are at risk in onboarding”
• AI generates: get_at_risk_onboarding(days_since_signup: number, limit: number)

• Prompt: “Create a tool to get recommended next steps for a customer in onboarding”
• AI generates: get_onboarding_next_steps(customer_id: string)

• Prompt: “Create a tool to analyze onboarding effectiveness and time-to-value”
• AI generates: analyze_onboarding_effectiveness(days_back: number, subscription_tier: string)

​

Step 4: Test and Publish

1. Test onboarding progress tracking
2. Verify at-risk customer identification
3. Validate next step recommendations
4. Publish tools
5. Connect to agent builder

​

Example Agent Interactions

​

Scenario 1: Check Onboarding Status

1. Calls get_onboarding_progress("abc_corp")
2. Responds:
   • “ABC Corp onboarding status: 50% complete
   • Completed: Account setup, Data import
   • Pending: Dashboard creation, Team invitation
   • Last active: 3 days ago
   • Status: In Progress
   • Recommended: Create your first dashboard”

​

Scenario 2: Identify At-Risk Customers

1. Calls get_at_risk_onboarding(30, 20)
2. Analyzes results:
   • “Found 12 at-risk customers:
     • TechCo: 25% complete, stalled on milestone 2, inactive 10 days
     • StartupXYZ: 0% complete, no activity for 14 days
     • …”

​

Scenario 3: Onboarding Effectiveness Analysis

1. Calls analyze_onboarding_effectiveness(90, null)
2. Analyzes results:
   • “Onboarding effectiveness (last 90 days):
     • Average time to complete: 12 days
     • Completion rate: 68%
     • At-risk rate: 18%
     • Most common blocker: Dashboard creation (32% of stalled customers)“

​

Outcomes

​

Onboarding Success

• Completion Rate: 25% increase in onboarding completion
• Time to Value: 30% reduction in time-to-value
• At-Risk Identification: 60% earlier identification of stuck customers
• Customer Engagement: 40% improvement in early-stage engagement

​

Efficiency Gains

• Automated Tracking: 80% reduction in manual tracking time
• Proactive Outreach: 3x increase in proactive customer outreach
• Personalized Guidance: 50% improvement in onboarding recommendations
• Team Productivity: 2x more customers managed per success manager

​

Data-Driven Insights

• Pattern Recognition: Identification of successful onboarding paths
• Bottleneck Analysis: Clear visibility into where customers get stuck
• Effectiveness Metrics: Continuous monitoring of onboarding performance
• Optimization Opportunities: Data-driven improvements to onboarding flow

​

Best Practices

1. Milestone Design: Define clear, measurable onboarding milestones
2. Regular Monitoring: Check onboarding progress daily
3. Proactive Outreach: Reach out to at-risk customers within 48 hours
4. Personalization: Tailor recommendations based on customer profile
5. Continuous Improvement: Use analytics to optimize onboarding flow

​

Next Steps

• Customer Support Agent Example - Build a support agent
• Customer Churn Predictor Example - Predict churn
• Product Usage Analyst Example - Analyze product usage