Lead Intelligence Platform

Problem Statement

Sales teams struggle with:

• Leads coming from multiple sources with no unified view
• Duplicated contact data polluting the CRM
• Missed follow-ups due to no intelligent prioritization
• CRM data sitting idle — not used for analytics or AI

This platform solves it by building a real-time, AI-powered lead management system that syncs Salesforce CRM data with Google services, scores leads automatically, and triggers smart follow-up actions.

---

High-Level Architecture

Salesforce Lead Created
        │
        ▼
Platform Event Published
        │
        ▼
Queueable Apex (async)
        │
        ▼ REST callout (OAuth2 / Named Credential)
Google Cloud Function
        │
        ├──► BigQuery (store lead data)
        │
        └──► Vertex AI (score lead)
                │
                ▼
        Callback → Salesforce REST Endpoint
                │
                ├──► Update AI_Score__c
                ├──► Auto-create Task
                └──► Google Calendar Event + Gmail Draft

---

Tech Stack

Salesforce

• Apex (Queueable, Batchable, REST)
• Platform Events
• Named Credentials (OAuth2)
• Custom Metadata (config)
• Lightning Web Components (dashboard)
• Custom Logging Framework

Google Cloud

• Cloud Functions (HTTP triggered)
• BigQuery (data warehouse)
• Vertex AI (lead scoring)
• Google Calendar API
• Gmail API

---

Project Structure

/salesforce
  /classes          → Apex classes
  /triggers         → Apex triggers
  /objects          → Custom object metadata
  /lwc              → Lightning Web Components
  /customMetadata   → Config records

/google-cloud
  /functions        → Cloud Function source code
  /bigquery         → Schema definitions
  /vertex           → AI model config

/docs
  /diagrams         → Architecture, sequence, ER diagrams
  /decisions        → Engineering decision records

---

Custom Fields on Lead Object

Field Label	API Name	Type	Purpose
AI Score	AI_Score__c	Number	Score returned by Vertex AI (0–100)
Sync Status	Sync_Status__c	Picklist	Pending / Synced / Failed
Duplicate Flag	Duplicate_Flag__c	Checkbox	Flagged by deduplication engine
External ID	External_Id__c	Text	Idempotency key for external systems
Retry Count	Retry_Count__c	Number	Tracks callout retry attempts
Last Sync Timestamp	Last_Sync_Timestamp__c	DateTime	When last synced to Google Cloud

---

Key Engineering Decisions

Why Queueable Apex over @future?

• Queueable supports chaining (job A → job B → job C)
• Supports complex object types as parameters
• Better error handling and monitoring via AsyncApexJob

Why Platform Events?

• Decouples trigger logic from external integration
• Survives transaction rollbacks (events are published even if DML fails)
• Enables microservice-style architecture inside Salesforce

Governor Limit Strategy

• All SOQL queries use Maps for O(1) lookups
• No SOQL or DML inside loops
• Bulk-safe: all code processes List not single records
• Callouts batched and chunked to respect limits

Idempotency

• Each outbound request carries a unique correlation_id
• Prevents duplicate processing if retried

---

Sprint Plan

Week	Focus
Week 1	Salesforce core — data model, trigger framework, async pipeline, logging
Week 2	Google Cloud — Cloud Function, BigQuery, simulated AI scoring
Week 3	Intelligence — dedup engine, Google Workspace, retry strategy
Week 4	UI, documentation, demo, polish

---

Setup Instructions

(Will be updated as each phase is completed)

1. Deploy Salesforce metadata using SF CLI
2. Configure Named Credential with Google Cloud OAuth2
3. Deploy Cloud Function to GCP
4. Update Custom Metadata with live endpoint URL
5. Assign permission sets to users

---