The phrase "two-minute underwriting" is the part of the announcement most likely to be misunderstood. To make it useful for someone with a term coming up, two things have to be separated: what underwriting actually involves, and what part of that workflow AI can realistically compress.
A 2026 analysis published by Sphera Credit synthesizing CMHC, FCAC, and big-bank data finds that mortgage underwriting in Canada typically takes 25 to 45 days from application to funding, with the underwriter's detailed review usually completing within 5 to 10 business days once the file is complete. Initial underwriter decisions on clean files often take 1 to 5 business days. The bulk of those calendar weeks is not decision time. It is document collection, verification, appraisal coordination, lawyer scheduling, and back-and-forth with the borrower.
The two-minute target compresses the decisioning piece of that workflow — the part where a file with verified inputs gets evaluated against a credit and qualifying-rate framework. It does not, on its own, eliminate the document collection, the appraisal, or the legal close. What it does is shrink the conditional-approval window enough that a renewer shopping a second quote can get a credible answer in minutes rather than days.
What Stays the Same — Including the Stress Test
The regulatory rails do not move because the decisioning got faster. OSFI's Guideline B-20 on residential mortgage underwriting practices applies to all federally regulated financial institutions engaged in residential mortgage lending, and it requires lenders to verify borrower income and ability to repay using a minimum qualifying rate above the contract rate. That framework — described in detail by OSFI's B-20 information sheet — is what an AI underwriting engine has to satisfy, not bypass.
A second guideline becomes more important as AI takes on more of the decisioning work. OSFI's updated Guideline E-23 on Model Risk Management (effective May 1, 2027) explicitly brings AI and machine-learning models into scope, requiring financial institutions to inventory, risk-rate, validate, and monitor models across the full lifecycle. That includes models used for underwriting and customer decisions, and the guideline requires institutions to address explainability, data quality, and bias for AI/ML systems. "Human oversight" in Nesto's framing is not just a comfort word. It is the operational answer to a regulatory framework that is sharpening at the same time the technology is being deployed.