Lending Application Agent
ExternalCopilot on Salesforce Agentforce automating small-business lending: identity verification, pre-qualification scoring with deterministic approval logic, document checklist generation, and application status tracking — zero human delay from inquiry to decision.
Overview
The Lending Application Agent automates the full small-business loan intake lifecycle on Salesforce Agentforce (ExternalCopilot, EinsteinServiceAgent planner). From first contact to pre-qualification decision, every step is deterministic: identity verification binds the applicant, a scoring engine evaluates revenue / time-in-business / loan amount against hard-coded lending criteria, and the agent returns an instant Qualified / Conditional / Declined decision with a tailored document checklist — no human bottleneck, no subjective judgment on routine applications.
Problem
Community and mid-market lenders lose applicants to friction. A typical small-business loan inquiry takes 3-5 business days just to get a preliminary yes/no because intake staff must manually collect financials, run them through spreadsheet scoring models, and compile document requests. Applicants abandon or go to fintechs that give instant answers. Meanwhile, loan officers spend 60% of their time on applications that will never qualify — time that should go toward relationship lending and complex credit decisions.
Approach
Four Agent Script topics with a linear intake flow: (1) Identity Verification — collects applicant email, calls VerifyApplicant Apex action (Contact → active Application → Account chain), binds applicant_id as a session guard for all downstream actions. (2) Pre-Qualification Scoring — collects annual revenue, years in business, and requested loan amount via guided conversation. A deterministic Apex scoring engine applies three-tier logic: Qualified (revenue ≥ 3× loan AND ≥ 2 years in business), Conditional (meets one criterion), Declined (meets neither). No LLM involvement in the credit decision — fully auditable, zero hallucination risk on lending compliance. Creates a Loan_Application__c record with score, tier, and timestamp. (3) Document Checklist — based on the qualification tier, generates a tailored checklist: Qualified applicants get the standard 6-doc package (tax returns, bank statements, P&L, balance sheet, business license, personal guarantee). Conditional applicants get the standard package plus additional collateral documentation. Declined applicants receive denial reasoning and next-step guidance. (4) Application Status — queries Loan_Application__c for existing applications, returns status, document completion percentage, and estimated timeline. Auto-lookup pattern: Status topic calls the action immediately on entry since applicant_id is already bound.
Tech Stack
Results
Four lending workflows fully automated: instant pre-qualification with deterministic three-tier scoring, tailored document checklist generation, real-time application status tracking, and identity-verified session management. Scoring engine is fully auditable with zero LLM involvement in credit decisions — critical for lending compliance. Eliminates the 3-5 day intake bottleneck that causes applicant abandonment. Designed for community banks and SBA lenders — directly addresses the #1 lending operations pain point: time-to-decision on routine applications.