AI Credit® Trials

Nine systematic R&D trials across 145 live commercial transactions exceeding $150M. The evolution from automation to augmentation — an industry first, 2020–2023.

Industry First 145 Transactions · ~$150M 9 Trials · 2020–2023 First Funded · Jan 2023

Research Documentation

Download

AI Credit® Trials Report

Complete 13-page research document detailing all nine trials, methodologies, and findings across 145 transactions totalling ~$150M.

Download

AVA White Paper

Technical white paper on implementing Accelerated Virtual Analyst systems for commercial credit secured by real estate and equipment.

Between 2020 and 2023, AVA Credit ran a series of structured trials applying generative AI to commercial lending operations. These nine experiments progressed from early automation pilots to advanced human-AI augmentation, ultimately producing AI-assisted credit writeups that were successfully submitted and approved for U.S. SBA 504 transactions.

What began as an exploration of AI’s role in risk assessment evolved into a rethinking of how financial professionals engage with credit data. The trials demonstrated that the future of commercial lending is not about replacing human expertise, but about amplifying it through conversational AI systems to enhance decision-making, efficiency, and professional judgment.

Evolution from Automation to Augmentation

Traditional AI Automation

Trials 1–6 · 2020–2022

Risk assessment simulators, machine learning credit management, fraud detection, predictive analytics, robotic process automation, and blockchain custody systems. Heavy mathematics proved challenging for finance professionals to integrate effectively.

Natural Language Processing

Trial 7 · 2022–2023

Document analysis using NLP to extract and interpret borrower information from bank statements, tax returns, and financial reports. First breakthrough toward conversational data interaction.

Generative AI Revolution

Trials 8–9 · 2022–2023

Generative AI for credit document creation and interactive AI advisors using Large Language Models. The game-changing breakthrough: skilled professionals could finally “talk to their data” through natural conversation.

Defining Moment

The Core Scientific Catalyst

In December 2021, our $3.1M equipment financing for Core Scientific’s AI infrastructure became the defining moment. Financing critical power equipment for OpenAI’s GPU clusters exposed us to the transformative potential of Generative AI, igniting our obsession with human-AI collaboration that would drive the breakthrough discoveries in Trials 7–9.

Starting in Q3 2022, we became the first commercial lender to submit AI-generated credit writeups for funded U.S. SBA transactions. Our AI-enhanced underwriting analysis for government-backed loans represents a historic first in the 75-year history of the SBA 504 program.

The trials proved that AI’s greatest value in financial services isn’t automation — it’s augmentation. By enabling natural language interaction with complex credit data, we discovered how to amplify human expertise rather than replace it, establishing the foundation for the Governed AI tools being built today.

Research Highlight

Trials 1–6: Traditional AI Automation Era

The first six trials explored conventional AI applications across core lending operations: risk assessment simulators, machine learning credit management, fraud detection systems, predictive default analytics, and robotic process automation. Each trial tested whether traditional AI could replace manual processes in commercial lending workflows.

The mathematics-heavy approach proved challenging for finance professionals to integrate effectively, revealing that AI’s true value wouldn’t come from automation alone. Trial 6’s blockchain document custody system exemplified this learning — while the technology solved real operational problems, humans struggled to meaningfully interpret the data that AI could analyse effortlessly.

This foundational period established that successful AI implementation required bridging the gap between AI capabilities and human workflows, setting the stage for the natural language breakthroughs in Trials 7–9.

Research Highlight

Trials 7–9: NLP and Generative AI Revolution

Supercharged by insights from financing Core Scientific’s AI training infrastructure, Trials 7–9 marked our breakthrough from traditional automation to human-AI augmentation. Trial 7 deployed Natural Language Processing for document analysis, enabling extraction and interpretation of complex borrower data from financial statements and tax returns.

Trials 8–9 introduced Generative AI and Large Language Models for credit document creation and interactive data analysis. This revolutionary period established the “talk to your data” paradigm, where skilled finance professionals could engage with complex credit workflows through natural conversation rather than rigid software interfaces.

These final trials produced the first AI-generated credit writeups submitted for funded SBA transactions — validating that the future of commercial lending lies in amplifying human expertise through conversational AI, not replacing it with automation. That conviction is the foundation of the Governed AI Workbench.

The First Funded — January 2023 →

Open Source Legacy

Following the precedent set by Google’s open-sourcing of “Attention Is All You Need,” we’ve made our AI Credit® Trials research publicly available. This foundational work in financial AI represents a mid six-figure private investment in operationalizing Generative AI for secured lending — knowledge that should benefit the entire industry.

The trials established the blueprint for human-AI collaboration in financial services. These findings continue to drive AVA Credit’s mission: building Governed AI tools that put the professional in control.