US supervisory model-risk guidance (SR 26-2)
United States banks · effective April 17, 2026Replaced the earlier SR 11-7 guidance on model risk management. Confirm how it applies to your use of AI with your model-risk team.
Industries · Financial services
AI your risk team can approve, on the numbers your CFO watches
Banks, insurers, lenders and asset managers run on documents, tight margins and regulators who want evidence. We build AI that moves a reported number and arrives with the evidence model-risk review asks for.
Sound familiar?
What we hear from financial services leaders
Each of these is a P&L problem with a governance condition attached. We solve both at once.
See where AI pays“We have a dozen AI pilots, and none of them has changed a number we report.”
“Model risk will not approve anything it cannot explain or test.”
“Our analysts still read every credit agreement, claim file or loan package by hand.”
“Finance, risk and the data team each define a customer differently.”
“A vendor model is in production, and nobody owns its monitoring.”
“The regulator's date is fixed, and our evidence is a slide deck.”
Where AI pays
Use cases we build and run
Banking and private credit
Credit agreements and amendments read into the loan book, with every change traced to its clause. KYC document extraction. Payment matching.
Insurance
Claims intake and triage, underwriting document review, and the evaluation evidence AI models need for insurance regulators.
Title and mortgage
Underwriting copilots, loan-file quality checks and title-package review, with AI governance programs for sellers and servicers.
Asset and wealth management
Operations efficiency, advisor productivity and data foundations that make reporting and AI agree.
Data foundations
Semantic layers and governed data products, so finance, risk and AI share one set of definitions.
AI governance evidence
Inventories, evaluation sets, monitoring and documentation that model-risk and compliance teams can review.
Governed from the first build
Evidence is a deliverable at every phase
We are not an auditor. We build the controls and the evidence your model-risk, compliance and audit teams ask for, so approval is a step, not a stall.
- 01AI Value AcceleratorLeadership accelerator
- Use cases ranked by return
- A risk tier for each one
- An entry in your AI inventory
- 02FDSForward Deployed Strategists (FDS)
- The KPI and its owner named
- A control agreed before the build
- An evaluation set from your own cases
- A measured result against the control
- 03FDEForward Deployed Engineers (FDE)
- Human approval at decision points
- Every output traced to its source
- Documentation for model validation
- Monitoring for drift and quality
- Named owners for every alert
- 04ScaleMultiple transformations
- An evidence pack kept current
- KPI readout informing the next step
What is driving demand
Rules that shape AI in banking
Fannie Mae Lender Letter LL-2026-04
United States, mortgage · in effect since August 2026Sellers and servicers must maintain an AI and machine learning governance framework.
EU AI Act, high-risk obligations
European Union · apply from December 2, 2027Includes AI used for credit scoring of individuals.
Client stories
Financial services work in production
Also: DealMaker, an AI conversational hub for an online capital-raising platform.
Your line of business is not listed? The approach is the same wherever money and data are regulated.
Talk to us
Find where AI pays in your institution
Tell us what you are working on, and we will bring people who know your industry.