NAFCO Finance: combining AI credit scoring with automated decisioning for rapid application growth

The challenge

NAFCO Finance had contracted with a 3rd Party AI provider for a custom credit model, but the score alone would not be the basis for decisions. It needed to be an enhancement to their existing underwriting workflow at the moment a decision was made. Together, this multifaceted approach could render approvals with structures that fit their guidelines or decline with supporting adverse action reasons that the staff could confidently rely upon.

The goal was to put AI credit scoring to work within a reliable automated decisioning process while increasing underwriting capacity. NAFCO Finance desired and was anticipating growth in application volume, but wanted to approve more loans without loosening standards or adding more staff. Much of the buying team’s work relied on judgment refined over years and held by experienced buyers. Encoding that judgment was difficult, and errors in an automated decision could prove costly if not done properly. NAFCO Finance needed evidence from live applications before turning the system loose to approve loans automatically.

What LAUNCHER built

LAUNCHER integrated the 3rd Party AI credit scoring with appTRAKER’s decisioning rules and autoBUILT loan structuring. The solution combined the model’s insights with NAFCO Finance’s underwriting criteria to support automated decisions. Each capability was introduced in phases and tested against live applications before the system assumed more decisioning authority.

Phase 1: Add intelligence without changing decisions. LAUNCHER first embedded the 3rd Party AI score and supporting reason factors in the existing underwriting workflow. Scores recalculated as deals were reworked, keeping the analysis aligned with the current structure for every applicant, including a co-borrower. Buyers still made every decision; the model added intelligence and adverse-action support without changing authority.

Phase 2: Run automated decisioning in parallel. Next, LAUNCHER introduced a recommendation-only decisioning pass. The system showed whether a deal would qualify for automated approval and which lender-defined criteria it met or missed, but it was still dependent on a buyer to manually render the decision. Buyers compared those recommendations with their own decisions on live applications. This validated the logic and showed how often dealer-submitted structures already fit—and whether automated restructuring was needed—before any further automation was introduced.

Phase 3: Automate deal structuring. Once the recommendation logic had been validated, LAUNCHER configured autoBUILT to restructure deals that did not initially fit, adjusting only permitted structure elements within NAFCO Finance’s program tolerances. Production applications continued to aid in the vetting of the system recommended structures. Rule logic continued to be refined with buyer feedback until reaching a point where offers aligned with buyers’ reasoning and compliant with guidelines. Within this phase, decisions remained in the hands of buyers.

Phase 4: Begin limited automated approvals. Only after the earlier phases had been tested did LAUNCHER activate automated approvals. The initial release covered a conservative portion of NAFCO Finance’s portfolio—the smallest and most straightforward deals—while defined application categories remained outside automation. Automatically approved applications routed to a dedicated review queue. Within weeks, NAFCO Finance’s results supported a substantial increase in the approval limit.

The result

In the first full month, automated decisioning accounted for 38% of NAFCO Finance’s approvals and 24% of all submitted applications. The following month, those figures held at 35% of approvals and 24% of applications despite a materially larger application base.

Application volume rose more than 50% in one month, with the following month tracking toward another 30% increase. NAFCO Finance absorbed both increases without hiring an additional buyer.

The outcome remained under NAFCO Finance’s control. Approval limits, criteria, tolerances and excluded application types are configurable and can be adjusted using live results. NAFCO Finance chose to raise the limit because the system’s recommendations were consistently aligning with their buyers’ judgment, program guidelines, and supported by NAFCO Finance’s performance data.