NEOEDGE AIAn inventory gap: tools, vendor features, and spreadsheets that may count as models have never been listed in one place.
← All AI decisions
E-23 evidence work · Federally regulated institutions
One evidence gap. Scoped to what you need.
OSFI’s Guideline E-23 on model risk management takes effect on May 1, 2027, and its definition of a model includes AI and machine learning. If you have a specific inventory, evidence, or ownership gap for AI in one area, I can help your team close it.
New to E-23? Read the plain guide for smaller lenders.
Sam Talasila Former Head of LLMs & Data Products, Wealthsimple
When it fits
A specific gap, not a general worry.
This work starts once your institution has confirmed E-23 applies and a risk owner can name the gap.
An evidence gap: the AI is in use, but there is no record of how it was tested, how it performs, or when a person reviews it.
An ownership gap: the business, technology and risk teams each assume someone else is responsible.
Scope
First-line work, done properly.
What the work can cover
For one defined area.
- An inventory of AI and models in the defined area
- Evidence of how each one is tested and how it performs on real cases
- Named owners, review points, and escalation paths
- Documentation your model-risk team can use
What it doesn’t cover
Roles I don’t take on.
- Formal independent validation or review of models
- A readiness claim for the whole institution
- Legal or regulatory interpretation
- Independent review of anything I helped design or build. My testing of that is development work.
How it starts
Four things to agree before any work begins.
- First
Confirm it applies.
Your institution confirms that E-23 applies to the area and the tools in question.
- Second
Name the owner.
A responsible risk or model-risk owner sponsors the work and can make decisions.
- Third
Define the gap.
We agree the one area and the specific inventory, evidence, or ownership gap to close.
- Fourth
Assign formal roles.
Your institution or a qualified specialist takes on interpretation, validation and assurance.
The offer
Evidence your model-risk team can use.
Start with one area. The inventory is a fixed-fee first step, from CAD 10,000, so you can see the work before committing to more.
What’s included
- Inventory of one areaEvery AI tool, vendor feature and spreadsheet that may count as a model, listed in one place.
- Evidence packHow each one is tested and how it performs on real cases.
- Owner and escalation mapNamed owners, review points, and who acts when something goes wrong.
- Committee summaryA short summary your risk committee or board can read.
My guarantee
Usable Evidence Guarantee
If your model-risk team finds the documentation isn’t usable for their work, I revise it until it is, at no extra cost.
This is first-line work. It isn’t validation, assurance, or a readiness opinion. I act as an external adviser and disclose any implementation interest.
Why now
E-23 takes effect on May 1, 2027. Inventories, evidence and reviews all take time, and your model-risk team needs room to review the work before that date.
At one lenderTesting a chatbot before launch caught a wrong figure in a source file before any staff member saw it, and the lender’s funder reports now check themselves against the current version.
Which gap would you close first?
In 30 minutes, tell me about the area and the gap. We’ll see whether scoped E-23 work, a Decision Sprint, or neither is the right fit.