Form fields: Full Name (Text, Required), Email Address (Email, Required), Phone Number (Text, Optional), Date of Birth (Date, Required), Employment Status (Dropdown: Salaried / Self-Employed / Business Owner / Unemployed, Required), Monthly Income (Number, Required), Loan Amount (Number, Required), Loan Purpose (Dropdown: Home / Car / Personal / Education / Business Loan, Required), PAN Number (Text, Required).

Trims all field keys to handle leading/trailing spaces, validates required fields, calculates age from DOB (throws error if outside 21–65), and computes DTI ratio using estimatedEMI = loanAmount × 0.02 then DTI = (EMI ÷ monthlyIncome) × 100.

const d = $input.first().json;
const data = {};
Object.keys(d).forEach(key => { data[key.trim()] = d[key]; });

const name   = data['Full Name']?.toString().trim();
const email  = data['Email Address']?.toString().trim();
const income = parseFloat(data['Monthly Income'] || 0);
const loan   = parseFloat(data['Loan Amount'] || 0);
const dob    = data['Date of Birth'];
const emp    = data['Employment Status'];
const purpose= data['Loan Purpose'];
const pan    = data['PAN Number']?.toString().trim();

const missing = [];
if (!name)   missing.push('Full Name');
if (!email)  missing.push('Email');
if (!income) missing.push('Monthly Income');
if (!loan)   missing.push('Loan Amount');
if (!dob)    missing.push('Date of Birth');
if (!pan)    missing.push('PAN Number');
if (missing.length > 0) throw new Error('Missing: ' + missing.join(', '));

const age = Math.floor((new Date() - new Date(dob)) / (365.25*24*3600*1000));
if (age < 21 || age > 65) throw new Error('Age must be 21-65. Got: ' + age);

const estimatedEMI = loan * 0.02;
const dtiRatio = parseFloat(((estimatedEMI / income) * 100).toFixed(2));

return [{ json: {
  applicant_name: name, email, phone: data['Phone Number'] || '',
  age, employment: emp, monthly_income: income, loan_amount: loan,
  loan_purpose: purpose, pan_number: pan, dti_ratio: dtiRatio,
  submitted_at: data['submittedAt'] || new Date().toISOString()
}}];

Simulates a CIBIL score for demo purposes. Starts at base 650, adjusts for income level (±60), DTI ratio (±80), and employment status (+20 salaried, −150 unemployed). Adds small random variation for realism. In production, replace this node with an HTTP Request to CIBIL/Experian using the applicant’s PAN number.

const d = $input.first().json;
let baseScore = 650;
if (d.monthly_income > 100000)      baseScore += 60;
else if (d.monthly_income > 60000)  baseScore += 30;
else if (d.monthly_income < 30000)  baseScore -= 50;

if (d.dti_ratio < 20)       baseScore += 40;
else if (d.dti_ratio > 50)  baseScore -= 80;
else if (d.dti_ratio > 35)  baseScore -= 30;

if (d.employment === 'Salaried')    baseScore += 20;
if (d.employment === 'Unemployed')  baseScore -= 150;

baseScore += Math.floor(Math.random() * 30) - 10;
const creditScore = Math.min(900, Math.max(300, baseScore));

return [{ json: { ...d, credit_score: creditScore,
  bureau_name: 'CIBIL (Simulated)', bureau_status: 'Active',
  checked_at: new Date().toISOString() }}];

System message

You are a senior loan pre-screening AI for SecureBank India.
Evaluate the applicant data and respond ONLY with a valid JSON object.
No markdown, no explanation outside the JSON.

JSON schema you must follow exactly:
{
  "decision": "approve" | "manual_review" | "reject",
  "risk_score": integer between 0 and 100 (higher = riskier),
  "summary": "2 sentence plain English explanation",
  "flags": ["concern 1", "concern 2"],
  "recommended_amount": number or null,
  "interest_band": "e.g. 8.5%-10.2%" or null
}

Decision rules:
- approve       : credit_score >= 720 AND DTI < 35% AND stable employment
- manual_review : credit_score 600-719 OR DTI 35-50% OR self-employed
- reject        : credit_score < 600 OR DTI > 50% OR unemployed
- Always reject if age < 21 or age > 65
- recommended_amount must not exceed 60x monthly income

User message / prompt

Applicant: {{ $json.applicant_name }}
Age: {{ $json.age }}
Employment: {{ $json.employment }}
Monthly Income: Rs {{ $json.monthly_income }}
Loan Requested: Rs {{ $json.loan_amount }}
Loan Purpose: {{ $json.loan_purpose }}
Credit Score: {{ $json.credit_score }}
DTI Ratio: {{ $json.dti_ratio }}%
PAN: {{ $json.pan_number }}

Reads the AI output field (tries output, text, response, message.content for n8n version compatibility), strips markdown fences, parses JSON, and merges structured fields back into the item.

const input = $input.first().json;
const raw = input.output || input.text || input.response
  || input.message?.content || input.choices?.[0]?.message?.content || '';

const cleaned = raw.replace(/```json/gi,'').replace(/```/g,'').trim();
let ai;
try { ai = JSON.parse(cleaned); }
catch(e) { throw new Error('AI did not return valid JSON. Raw: ' + raw); }

return [{ json: { ...input,
  decision: ai.decision, risk_score: ai.risk_score,
  ai_summary: ai.summary,
  flags: Array.isArray(ai.flags) ? ai.flags.join(', ') : '',
  recommended_amount: ai.recommended_amount || null,
  interest_band: ai.interest_band || null
}}];

All three nodes POST to https://api.resend.com/emails with Header Auth → Resend-API. Each sends a different HTML email body with n8n expressions injecting applicant data.

The Merge node (Append mode) reunites all three branches. The Postgres node then inserts the complete record. The query uses RETURNING id so the application ID is available for downstream use.

INSERT INTO loan_applications (
  applicant_name, email, phone, age, employment,
  monthly_income, loan_amount, loan_purpose,
  credit_score, dti_ratio, decision, risk_score,
  ai_summary, flags, processed_at
) VALUES (
  '{{ $json.applicant_name }}', '{{ $json.email }}',
  '{{ $json.phone }}', {{ $json.age }},
  '{{ $json.employment }}', {{ $json.monthly_income }},
  {{ $json.loan_amount }}, '{{ $json.loan_purpose }}',
  {{ $json.credit_score }}, {{ $json.dti_ratio }},
  '{{ $json.decision }}', {{ $json.risk_score }},
  '{{ $json.ai_summary }}', '{{ $json.flags }}', NOW()
) RETURNING id;