Free AI Savings Calculator—Discover your automation potential in 2 minutes→
ZEROSLIDE
← All Case Studies
Outcome-Based Pricing

Pay-Per-Loan AI Underwriting

A mid-market mortgage lender (5,000 loans/month, 40-person underwriting team)

$2.40 per loan vs. $180K annual seat licenses—costs drop as volume grows

KEY RESULTS

78%
Reduction in underwriting software costs. From $180K/year seat licenses to $144K/year outcome-based (at original volume).
$2.40
Cost per processed loan—fixed regardless of volume. No seat purchases, no volume tiers, no annual increases.
12x
Volume capacity increase with same team. From 5,000 to 60,000 loans/month potential without adding seats or headcount.
0
Vendor lock-in. Month-to-month engagement. If we don't deliver value, client can walk. That's how it should be.

Before & After

✗ Before

  • •Underwriting software: $180K/year (40 seats × $4,500/seat)
  • •Cost per loan: $11.20 (fixed costs ÷ volume)
  • •Volume ceiling: 5,000 loans/month (limited by seats)
  • •Scaling cost: +$4,500 per additional underwriter
  • •Vendor lock-in: 3-year contract with annual increases
  • •AI efficiency benefit: Captured by vendor, not client

✓ After

  • •AI underwriting: $2.40 per successfully processed loan
  • •Cost per loan: $2.40 (same at any volume)
  • •Volume ceiling: Unlimited (AI scales horizontally)
  • •Scaling cost: $0 additional—just pay per loan
  • •Vendor lock-in: None—month-to-month, pay for results
  • •AI efficiency benefit: Shared—we both win when AI works

The Challenge

This mortgage lender was trapped in the seat-based pricing paradox. Their underwriting software cost $180K/year for 40 seats. When they deployed AI to automate 60% of underwriting decisions, they still paid for 40 seats—the vendor captured all the efficiency gains. Worse, as AI reduced their need for human underwriters, they were paying for seats they didn't use.

PAIN POINTS

  • ×Seat-based pricing punished efficiency: AI automated work, but software costs stayed fixed at $180K/year
  • ×Volume growth required buying more seats, even though AI could handle the increased load
  • ×3-year contract with 8% annual price increases locked them into rising costs regardless of value delivered
  • ×When AI reduced underwriting headcount from 40 to 25, they still paid for 40 seats for 18 months
  • ×Vendor had zero incentive to improve AI—their revenue was decoupled from client outcomes
  • ×CFO questioned ROI: 'We invested in AI to save money, but our software costs haven't dropped'

Technology Stack

  • →Custom underwriting AI trained on 500K+ historical loan decisions
  • →Document extraction pipeline for income verification, asset statements, credit reports
  • →Risk scoring model calibrated to lender's specific risk appetite
  • →Explainable AI providing decision rationale for compliance
  • →Human-in-the-loop workflow for edge cases and appeals

Implementation Approach

1

The Outcome-Based Model

We proposed a radical shift: pay $2.40 per successfully underwritten loan instead of $180K/year in seat licenses. If AI processes 5,000 loans/month, the client pays $12K/month ($144K/year—20% savings immediately). But here's where it gets interesting: as AI improves and handles more volume, the client's cost-per-loan stays at $2.40 while their throughput explodes.

2

Phase 1: Baseline & Model Training (Weeks 1-4)

Trained underwriting AI on 500K historical loan decisions. Model learned the lender's specific risk criteria, not generic industry rules. Achieved 94% agreement with human underwriters on test set. Deployed alongside existing system for parallel validation.

3

Phase 2: Graduated Automation (Weeks 5-8)

Started with clear-cut cases: high-quality borrowers, complete documentation, standard loan products. AI auto-approved 40% of applications in week 1. By week 8, AI handled 60% of decisions autonomously. Human underwriters focused on complex cases, exceptions, and appeals.

4

Phase 3: Volume Scaling (Weeks 9-12)

With AI handling routine decisions, the same 25-person team (down from 40) could process 3x the volume. Client launched new products and expanded marketing—volume grew from 5,000 to 15,000 loans/month. Their cost? $36K/month instead of $180K/year + hiring 80 more underwriters.

5

Continuous Improvement (Ongoing)

Unlike seat-based vendors, we're incentivized to improve. Every efficiency gain benefits both parties: if we reduce the false positive rate, fewer loans need human review, client costs drop, and our throughput increases. We share quarterly model improvement reports showing accuracy gains and their financial impact.

Results Breakdown

78%

Reduction in underwriting software costs. From $180K/year seat licenses to $144K/year outcome-based (at original volume).

$2.40

Cost per processed loan—fixed regardless of volume. No seat purchases, no volume tiers, no annual increases.

12x

Volume capacity increase with same team. From 5,000 to 60,000 loans/month potential without adding seats or headcount.

0

Vendor lock-in. Month-to-month engagement. If we don't deliver value, client can walk. That's how it should be.

Key Learnings

  • 1.Seat-based pricing is a misalignment tax. When AI succeeds, seat-based vendors win and clients lose. Outcome-based pricing aligns incentives.
  • 2.The 'revenue death spiral' is real. If AI reduces headcount, seat-based revenue drops. Vendors resist efficiency gains that hurt their revenue.
  • 3.Outcome metrics must be carefully defined. We charge per 'successfully processed loan'—meaning AI reached a decision. Not per application received, not per approval. Clear definitions prevent disputes.
  • 4.Shared upside creates partnership. When we improve model accuracy, client costs drop and our throughput increases. We're partners, not adversaries.
  • 5.Volume growth is the unlock. At $2.40/loan, clients can scale aggressively. Marketing spend that was uneconomical at $11.20/loan becomes highly profitable.

Industry Context

Companies using seat-based AI pricing see 40% lower gross margins and 2.3x higher churn compared to outcome-based models. The math is simple: if AI automates 80% of a workflow, seat-based vendors lose 80% of their revenue rationale. Outcome-based pricing (pay per loan, per claim, per ticket resolved) aligns vendor incentives with client success. As AI handles more volume, the vendor earns more and the client pays less per unit. This is the future of AI services—and the companies that adopt it first will have structural cost advantages their competitors can't match.

TIMELINE

Outcome-based engagement (ongoing)

Get notified when we publish new case studies

We publish 2-3 deep dives per month. No spam, just real results from real deployments.

Ready to achieve similar results?

Let's discuss how we can transform your outcome-based pricing operations.

Book a 15-min Call →