Mortgage underwriting conditions are the final bottleneck before clear-to-close. Each loan file generates 10-25 conditions — requests for additional documents, explanations, or verifications. Manual clearing takes 2-4 hours per file and stretches across 5-10 calendar days. Industry leaders like Gateless report 70-75% automated condition clearing rates. Confer's architecture — 8 specialized AI agents working across 32+ MCP tools spanning documents, income, assets, credit, and compliance — pushes beyond 70% by automating not just document matching, but verification, calculation, and validation workflows.
What Are Mortgage Underwriting Conditions?
After initial underwriting review, the underwriter generates a list of conditions — items that must be satisfied before the loan can close. Think of conditions as the underwriter's checklist for risk mitigation. Typical conditions include:
Document Conditions
- • Provide pay stubs for last 30 days
- • Submit complete tax returns for 2024-2025
- • Provide proof of homeowner's insurance
- • Update credit report (older than 120 days)
Verification Conditions
- • Verify employment and income
- • Verify source of deposit ($5,000 on 10/15)
- • Verify rental history for past 2 years
- • Verify outstanding judgments satisfied
Explanation Conditions (LOEs)
- • Explain gap in employment (3/2023-7/2023)
- • Explain credit inquiry from Auto Lender Inc
- • Explain 30-day late payment on Visa 12/2024
- • Explain self-employment income decline
Calculation Conditions
- • Recalculate DTI with updated income
- • Verify reserves meet 6-month requirement
- • Calculate rental offset for investment property
- • Verify gift funds seasoned per guidelines
A typical loan file has 10-25 conditions after initial underwriting. Clearing them all is the final gate before clear-to-close — and it's where loans often stall for days or weeks.
The Manual Condition Clearing Workflow
Here's what happens manually for each condition:
Processor reads condition text, determines what document or action is needed.
Search eFolder, email attachments, or borrower portal for the required document.
Check that document satisfies the specific requirement (date range, amount, signature, etc.).
Write notes explaining how condition was satisfied. Attach document reference.
Update LOS status. Notify underwriter if needed.
Multiply this across 100 loans per month: 200-400 hours of processor labor just clearing conditions. That doesn't include time spent chasing borrowers for missing documents or re-reviewing corrected submissions.
The 70% Benchmark: What It Means
Industry Auto-Clearing Rates
Vendors like Gateless pioneered AI-powered condition recognition and claim 70-75% automated clearing rates. Here's how it works:
- AI parses condition text: "Provide pay stubs for last 30 days"
- When borrower uploads documents, AI classifies: "This is a pay stub"
- AI checks date range: pay stub dated within last 30 days
- If match confidence > threshold: auto-clear condition
This approach works well for straightforward document-matching conditions. But 70% cleared means 30% still require manual intervention. For a file with 20 conditions, that's 6 conditions per loan that processors still handle manually — roughly 30-60 minutes of labor.
The remaining 30% are where complexity lives: income calculations that require trending analysis, asset verifications needing 2-month bank statement reviews, explanations requiring borrower LOEs, and compliance checks spanning multiple data sources. Simple document matching can't automate these — you need specialized intelligence for each domain.
Confer's Approach: 8 AI Agents, 32+ MCP Tools
Confer doesn't use a single condition-matching algorithm. Instead, 8 specialized AI agents collaborate to handle different aspects of loan processing. Each agent has access to domain-specific MCP (Model Context Protocol) tools:
Document Agent
MCP Tools: classify_document, extract_fields, validate_completeness, check_signatures
Handles all document-based conditions. Classifies uploads, extracts data, validates date ranges and signatures. Auto-clears conditions like "Provide W-2" when valid W-2 is uploaded and verified.
Income Agent
MCP Tools: calculate_income, verify_employment, trending_analysis, validate_paystubs
Auto-clears income verification conditions. Runs instant VOE, calculates qualifying income per Fannie Mae 1084, detects declining trends, flags variances. Clears "Verify employment and income" without processor intervention.
Asset Agent
MCP Tools: verify_assets, check_reserves, large_deposit_analysis, source_of_funds
Analyzes bank statements for reserve calculations, identifies large deposits requiring explanation, verifies gift fund documentation, calculates 2-month average balances. Auto-clears asset conditions when guidelines are met.
Credit Agent
MCP Tools: pull_credit, analyze_tradelines, verify_liabilities, check_inquiries
Handles credit-related conditions: pulls updated credit reports when needed, validates tradeline payments, explains inquiries, verifies debt payoffs. Clears "Update credit report" condition by pulling fresh report and validating no material changes.
Compliance Agent
MCP Tools: qm_atr_check, trid_timers, hmda_validation, dual_wire_verification
Automates compliance-related conditions. Validates QM/ATR 8 factors, monitors TRID timers, verifies dual wire transfers over $500K, populates HMDA fields. Clears compliance conditions with full audit trail.
AUS Agent
MCP Tools: run_aus, parse_findings, export_mismo, validate_eligibility
Runs Desktop Underwriter / Loan Product Advisor, exports MISMO 3.4, parses findings, validates loan eligibility. Auto-clears conditions requiring AUS re-run when data changes (updated income, new credit report).
Condition Agent
MCP Tools: parse_conditions, route_to_agent, track_status, validate_clearance
Orchestrator that parses condition text, routes each condition to the appropriate specialist agent, tracks clearing status, validates that all required elements are satisfied before marking as cleared.
Underwriting Agent
MCP Tools: risk_assessment, guideline_check, exception_analysis, generate_conditions
Generates initial conditions based on underwriting review. Works with other agents to validate that cleared conditions actually satisfy guideline requirements. Final authority on whether condition resolution is acceptable.
This multi-agent architecture means Confer doesn't just match documents to conditions — it performs the actual verification, calculation, and validation work that would otherwise require processor expertise.
Comparing Manual vs. Industry vs. Confer
| Metric | Manual (Baseline) | Industry (70% Auto) | Confer (Multi-Agent) |
|---|---|---|---|
| Auto-clearing rate | 0% | 70-75% | 85-90% |
| Time per 20-condition file | 2-4 hours | 30-60 min (6 manual) | 10-20 min (2-3 manual) |
| Document matching | Manual | ✓ Automated | ✓ Automated |
| Income verification | Manual | Partial | ✓ Full automation |
| Asset/reserve calculations | Manual | Manual | ✓ Automated |
| Compliance checks | Manual | Manual | ✓ Automated (QM/ATR, TRID) |
| Audit trail quality | Notes-based | Document references | Full MCP tool logs |
| Calendar days to clear-to-close | 5-10 days | 3-5 days | 1-2 days |
The Hybrid Approach: When to Auto-Clear, When to Flag
Not every condition should be auto-cleared. Confer uses confidence scoring to determine the appropriate action:
High Confidence (>85%)
Action: Auto-clear
Standard document matches, instant VOE success, credit report updates with no material changes, deterministic calculations. These clear immediately with full audit trail.
Medium Confidence (50-85%)
Action: Flag for review
Ambiguous documents, partial matches, calculations with edge cases. Agent provides suggested resolution and confidence score. Processor reviews and approves/rejects in seconds.
Low Confidence (<50%)
Action: Human required
Subjective explanations (LOEs), complex income scenarios, unusual asset sources. AI cannot safely resolve these without expert judgment. Routed to processor immediately.
This confidence-based routing means Confer auto-clears more conditions than simple document matching (pushing past 70%) while maintaining quality and compliance standards higher than manual review.