AVA Credit Inc.
Research
Human-AI R&D in financial services since 2017. Building Governed AI tools for licensed mortgage professionals in the era of OSFI E-23 and B-10.
The Journey
AVA’s AI Research Journey
From early conviction to Governed AI, 2017 to now.
Early recognition of transformer architecture’s potential in financial services. Filed AI Credit® trademark in October 2017, two months after encountering Attention Is All You Need. Trademark registered 2019.
Nine systematic R&D trials across 145 live commercial transactions totalling ~$150M. January 2023: the historic milestone. The first known AI-generated credit writeup submitted, approved, and funded through the 75-year-old U.S. SBA 504 program.
AVA incorporated in Delaware (March 2023), redomiciled to Toronto as AVA Credit Inc. (May 2024). RAG experiments and lessons learned. The name AVA, Accelerated Virtual Analyst, coined in early 2023. Research papers published.
2025: Perplexity AI Business Fellowship. Enterprise contract with TSX: DLCG Dominion Lending Centres, Canada’s first enterprise AI training pilot for mortgage professionals. Isaac AI agent deployed at dlcg.io. 2026: Governed AI tools for Canadian financial services under OSFI E-23 and B-10. Project Workbench in active development.
The AVA Story
AVA wasn’t built from a pitch deck. Eight years of experimentation at the intersection of AI and financial services, live transactions, failed experiments, and a fellowship alongside the CEOs building the frontier, shaped a clear conviction: the most durable AI tools in regulated industries are Governed ones.
What We Build On
AVA Research began in 2020 with a simple question: how do experienced professionals actually work with AI to get better outcomes? Not in theory, but in practice, under real conditions, with real stakes.
Over four years we conducted systematic trials across 145 commercial transactions totalling ~$150M. We documented what worked, what failed, and why. That body of knowledge is the foundation Project Workbench is built on.
OSFI’s E-23 guideline and B-10 third-party risk framework are reshaping how AI tools can be deployed in regulated financial services. Every tool inside Project Workbench is assessed against these frameworks, with logging, inventory, human-in-loop controls, and audit-ready documentation.
The research output flows directly into the Workbench. R&D validates the tools. The Workbench delivers them. That loop, build, validate, deploy, learn, is what makes Governed AI practical rather than theoretical.
Research Areas