Sales Assistant

​

Overview

​

What the Agent Needs to Accomplish

• Analyze sales pipeline and deal stages
• Identify high-value opportunities
• Forecast revenue based on pipeline data
• Track deal progress and identify stalled deals
• Provide insights on win rates and conversion metrics
• Recommend next actions for deals
• Analyze sales performance by rep, region, or product

​

How Pylar Helps

• Unified Pipeline View: Combining CRM data, sales activity, and historical performance
• Real-time Analysis: Querying current pipeline status and forecasts
• Multi-Source Integration: Joining CRM data with product catalog and customer data
• Pattern Recognition: Identifying successful deal patterns
• Actionable Insights: Generating recommendations based on data

​

Without Pylar vs With Pylar

​

Without Pylar

• ❌ Multiple systems (CRM, sales tools, analytics)
• ❌ Complex API integrations for each system
• ❌ Manual pipeline analysis and forecasting
• ❌ Difficult to correlate deals with customer data
• ❌ Time-consuming report generation
• ❌ Limited real-time insights

• 3-4 different API integrations
• Custom pipeline analysis logic
• Manual forecasting calculations
• Complex data correlation
• ~4-5 weeks development time

​

With Pylar

• ✅ Single endpoint for all sales data
• ✅ Real-time pipeline analysis
• ✅ Automated forecasting
• ✅ Unified customer-deal view
• ✅ Easy to update analysis logic
• ✅ Built-in analytics on sales patterns

• Connect 3-4 data sources (1 hour)
• Create sales pipeline views (2.5 hours)
• Build MCP tools with AI (1.5 hours)
• Connect to agent builder (15 minutes)
• Total: ~5 hours

​

Step-by-Step Implementation

​

Step 1: Connect Data Sources

1. Connect CRM (Salesforce/HubSpot - deals, opportunities, contacts)
2. Connect Sales Activity (Email, calls, meetings tracking)
3. Connect Product Catalog (Product pricing, configurations)
4. Connect Customer Data (Account information, historical purchases)

​

Step 2: Create Sales Pipeline Views

CREATE VIEW pipeline_overview AS
SELECT 
  d.deal_id,
  d.deal_name,
  d.account_name,
  d.owner_name,
  d.stage,
  d.amount,
  d.probability,
  d.expected_close_date,
  d.created_date,
  DATEDIFF(d.expected_close_date, CURRENT_DATE) as days_to_close,
  -- Weighted pipeline value
  d.amount * (d.probability / 100) as weighted_value,
  -- Deal age
  DATEDIFF(CURRENT_DATE, d.created_date) as deal_age_days,
  -- Activity metrics
  a.email_count,
  a.call_count,
  a.meeting_count,
  a.last_activity_date,
  DATEDIFF(CURRENT_DATE, a.last_activity_date) as days_since_last_activity,
  -- Risk indicators
  CASE 
    WHEN DATEDIFF(CURRENT_DATE, a.last_activity_date) > 14 THEN 1
    ELSE 0
  END as stalled_flag
FROM crm.deals d
LEFT JOIN sales.activity_summary a ON d.deal_id = a.deal_id;

CREATE VIEW revenue_forecast AS
SELECT 
  DATE_TRUNC('month', expected_close_date) as forecast_month,
  stage,
  COUNT(*) as deal_count,
  SUM(amount) as total_pipeline_value,
  SUM(amount * (probability / 100)) as weighted_forecast,
  AVG(probability) as avg_probability,
  SUM(CASE WHEN stage = 'Closed Won' THEN amount ELSE 0 END) as closed_won_value
FROM pipeline_overview
WHERE expected_close_date >= CURRENT_DATE
GROUP BY DATE_TRUNC('month', expected_close_date), stage
ORDER BY forecast_month, stage;

CREATE VIEW opportunity_analysis AS
SELECT 
  p.*,
  -- Win probability based on stage
  CASE 
    WHEN p.stage = 'Closed Won' THEN 100
    WHEN p.stage = 'Negotiation' THEN 75
    WHEN p.stage = 'Proposal' THEN 50
    WHEN p.stage = 'Qualified' THEN 25
    ELSE 10
  END as stage_based_probability,
  -- Recommendation
  CASE 
    WHEN p.stalled_flag = 1 THEN 'Re-engage customer'
    WHEN p.days_to_close < 7 AND p.probability < 50 THEN 'Increase activity'
    WHEN p.weighted_value > 100000 THEN 'Prioritize - high value'
    ELSE 'Continue normal cadence'
  END as recommended_action
FROM pipeline_overview p;

​

Step 3: Create MCP Tools with AI

• Prompt: “Create a tool to get sales pipeline summary by stage, rep, or date range”
• AI generates: get_pipeline_summary(stage: string, rep_name: string, start_date: date, end_date: date)

• Prompt: “Create a tool to forecast revenue for a specific time period”
• AI generates: forecast_revenue(months_ahead: number, include_probability: boolean)

• Prompt: “Create a tool to identify high-value opportunities or stalled deals”
• AI generates: identify_opportunities(min_value: number, stage: string, stalled_only: boolean)

• Prompt: “Create a tool to analyze sales performance by rep, region, or product”
• AI generates: analyze_sales_performance(group_by: string, start_date: date, end_date: date)

​

Step 4: Test and Publish

1. Test pipeline analysis
2. Verify revenue forecasting
3. Validate opportunity identification
4. Publish tools
5. Connect to agent builder

​

Example Agent Interactions

​

Scenario 1: Pipeline Summary

1. Calls get_pipeline_summary(null, null, null, null)
2. Responds:
   • “Current Pipeline Status:
     • Total Pipeline Value: $2.4M
     • Weighted Forecast: $1.2M
     • Deals by Stage:
       • Qualified: 45 deals ($800K)
       • Proposal: 28 deals ($900K)
       • Negotiation: 12 deals ($500K)
       • Closed Won this month: $200K”

​

Scenario 2: Revenue Forecast

1. Calls forecast_revenue(3, true)
2. Responds:
   • “Q2 Revenue Forecast:
     • Best Case: $1.8M (all deals close)
     • Weighted Forecast: $1.2M (probability-adjusted)
     • Committed: $450K (closed deals)
     • At Risk: $200K (stalled deals)“

​

Scenario 3: Opportunity Identification

1. Calls identify_opportunities(50000, null, true)
2. Responds:
   • “High-Value Stalled Opportunities:
     • Enterprise Corp: $150K, stalled 18 days, recommended: Re-engage
     • TechStartup: $85K, stalled 12 days, recommended: Schedule meeting
     • …”

​

Outcomes

​

Sales Performance

• Pipeline Visibility: 100% real-time pipeline visibility
• Forecast Accuracy: 25% improvement in forecast accuracy
• Deal Velocity: 20% faster deal progression
• Win Rate: 15% improvement in win rates

​

Efficiency Gains

• Report Generation: 90% reduction in report creation time
• Opportunity Identification: 3x faster identification of high-value deals
• Activity Tracking: Automated tracking of sales activities
• Team Productivity: 2x more deals managed per rep

​

Data-Driven Decisions

• Pattern Recognition: Identification of successful deal patterns
• Risk Identification: Early detection of at-risk deals
• Performance Insights: Clear visibility into rep and region performance
• Optimization: Data-driven improvements to sales process

​

Best Practices

1. Regular Updates: Keep pipeline data current
2. Activity Tracking: Monitor sales activity regularly
3. Forecast Reviews: Review forecasts weekly
4. Stalled Deal Follow-up: Follow up on stalled deals within 48 hours
5. Performance Analysis: Analyze performance trends monthly

​

Next Steps

• Revenue Operations Agent Example - Track revenue operations
• Lead Qualification Agent Example - Qualify leads automatically
• Customer Support Agent Example - Support customers effectively