About
The work behind Ava predates the first GPT.
AVA Credit Inc. is a Canadian federal corporation in Toronto. It builds one product, Ava: governed AI for Canadian SMEs and independent professionals who hold sensitive client files.
Ava started with a trademark and an operational mission. In December 2017, a commercial lender read a research paper about a new kind of neural network and filed a trademark on what it would mean for credit. Everything below followed from that.
The AVA Timeline
AI Credit® trademark filed.
December 2017, the month Attention Is All You Need was presented at NeurIPS.
AI Credit® trademark registered.
Registered October 2019, TMA1,056,883.
AI Credit® trials.
Nine R&D trials across 145 commercial transactions, ~$150M.
First funded.
First known Human-AI enabled credit writeup approved in the history of the U.S. SBA 504 program.
Incorporated in Toronto.
AVA Credit Inc. domiciled in Canada’s financial capital.
JPMorgan platform selection.
Selected for the JPMorgan equity management platform.
Perplexity AI Business Fellowship.
Inaugural cohort with CEOs from Perplexity, NVIDIA, Databricks, Box, and Bridgewater.
AI Essentials for DLCG.
A pioneering enterprise AI training program for DLCG Mortgage Group (TSX: DLCG), a network of 9,000+ mortgage professionals.
One product.
AVA Credit Inc. is organized entirely around Ava.
Live 1 November.
Ava 1.369 in use at our first client, Torcal Credit Inc.
Advance sales.
Development partners join in January. Advance sales with partners and new customers from Q2 2027.
Public release.
Generation 1 public, and Generation 2 live, from 1 July 2027.
Today
One product, one market, and one rule about client data.
Ava is an accelerated virtual analyst, and it is the governance layer. It gives firms the best AI models on their clients’ files, with identity kept in the firm and every use on the record.
Every experiment, including the ones we abandoned, led back to one boundary: the client’s identity stays with the firm. Ava holds that boundary. It removes names and numbers before a file goes to the AI, puts them back when the work returns, and keeps a Governed AI record of every step.
What it does
Two jobs. De-identify and restore, so the work is done in the best AI available, frontier or local, and the AI model never sees who the client is. And the Governed AI record: model inventory, usage log, data governance and human review, kept on every file.
Who it is for
The five-lawyer practice. The ten-professional accounting firm. The finance team of a mid-sized manufacturer. The sole licensed mortgage broker. Bound to confidentiality, without a bank’s capacity to build around it.
How it is built
In Toronto, with Canada’s regulators in view, federal and provincial, and the duties they share: privacy, confidentiality and a record of how AI was used. One rule in every generation, and the boundary widens: the professional’s computer, then the firm, then the firm’s own model.
Founder
Matthew Dobson
Founder and Chief Executive Officer
2025 Perplexity AI Business Fellow
A commercial lending veteran of twenty-five years across GE Capital, TD Bank and institutional private credit, most recently as Chief Credit Officer for a bank-backed private credit fund.
He was among the first in financial services to read what transformer architecture would mean for regulated lending work, and to file on it. He developed the RGI Framework — Retrieve, Generate, Interact — for human-AI collaboration, and has trained 2,271 licensed professionals across 29 live sessions at Canada’s largest publicly traded mortgage brokerage network, DLCG Mortgage Group (TSX: DLCG).
Developer
AVA Credit Inc. builds Ava.
Toronto, Ontario M5H 1J9