Workflow Automation · Portfolio Case Study
An end-to-end automation that captures deal intelligence from sales calls, runs it through AI-powered risk analysis, routes for human approval where needed, and delivers a complete customer context package to Zendesk and Salesforce — before the first onboarding touch.
System Architecture
Context
How It Works
The workflow begins when a sales rep signals call completion. Gong's native integration pushes the call record to Salesforce, where the transcript is pulled.
The raw transcript is cleaned, key fields are extracted (account, rep, deal stage, promises), and a structured payload is assembled for the AI layer.
The AI generates a deal summary, identifies promises and commitments made, assigns a confidence score, and classifies the deal as High Risk or Low Risk.
Financial promises, discounts, unclear outcomes, or low-confidence scores route to dedicated Slack channels (#finance-approvals, #sales-approvals, #cs-approvals). Approvers can approve, reject, or edit inline.
Low-risk deals bypass approval and push directly. Approved deals follow the same path. Rejected deals trigger Salesforce logging, rep notification, and a manual follow-up task — without touching Zendesk.
The CS team sees full deal context, promises made, and risk flags before the first customer touch. An audit trail and approval log are maintained throughout.
Approval/rejection patterns are tracked. False positives and negatives inform threshold adjustments. Risk flags and confidence scoring improve with usage.
Design Outcomes
Manual handoff documents required from sales reps post-call
Approval routing: finance, sales, and CS exceptions handled separately
Deals arrive in Zendesk with structured context before first CS touch
Tech Stack