NEOEDGE AIE-23 model inventory template. Free, built from OSFI’s own list.
OSFI’s Guideline E-23 lists, in its Appendix 1, the facts to keep for each model. Six apply to every model you identify. Eleven more apply to models with non-negligible risk, which go in your model inventory. This free spreadsheet has a column for each, plus two filled-in examples.
Four tabs: Read me, All models, Model inventory, Change log.
Which fields apply to every model?
OSFI expects you to keep six facts for each model you identify, at a minimum: a model ID, its name with a description of its key features and use, its risk rating, its owner, its developer, and its origin, such as built in-house or bought from a vendor.
Which extra fields go in the model inventory?
For models with non-negligible risk, OSFI adds eleven: version, date it went into production, reviewer, approver, dependencies, data sources with a description, approved uses, limitations, date of the most recent review, monitoring status, and next review date.
How do I fill in the risk rating?
OSFI asks for clear, measurable criteria that mix numbers and judgment. Its examples include the size and growth of the portfolio a model covers, financial or operational impact, model complexity or autonomy, data reliability and customer impact. A rating can start as provisional and be confirmed at review or approval. Vendor models are rated on their own. The template uses four levels as an example; OSFI doesn’t set the scale.
Can a model be rated negligible?
Yes. OSFI lets you use a rating that means negligible risk, which exempts a model from the full lifecycle rules. OSFI still expects “a robust process to approve and track such exemptions.” Those models stay on the All models tab with their six facts.
What does a filled-in row look like?
The template has two made-up examples. One is an in-house mortgage pricing spreadsheet. The other is a vendor AI tool that reads income documents. Both are rated, owned and dated, and the vendor tool is rated on its own, as E-23 asks.
| Model ID | Name and use | Risk rating | Owner | Developer | Origin |
|---|---|---|---|---|---|
| M-001 | Residential mortgage pricing sheet. Sets offered rates from the cost of funds and risk add-ons for each product | Medium (provisional) | Head of Residential Lending | Treasury analyst | In-house |
| M-002 | Vendor income-document reader. An AI tool that reads pay stubs and T4s and fills in income fields for underwriters | High (provisional) | Head of Underwriting | Vendor | Vendor |
Is a spreadsheet good enough?
The guideline doesn’t name a system. It asks for an inventory kept at the enterprise level that is “accurate, evergreen, and subject to robust controls” and updated promptly. A spreadsheet can work for a small inventory if you control who edits it and keep its change history. The template has a change log tab for that.
How do I find the models in the first place?
OSFI expects you to survey the institution from time to time for models, including vendor and third-party ones. You also track models retired recently. Ask each team which spreadsheets, vendor tools and AI tools turn data into a result someone acts on. The plain guide explains what counts as a model.
Where I can help
I build the first inventory for one area of your business, with risk ratings, owners and first-line documentation your model-risk team can use. Fixed fee from CAD 10,000. This is first-line work. It isn’t validation, assurance or a readiness opinion.
More on E-23 and AI
- E-23 for smaller lenders: a plain guide
- Validation vs verification under E-23
- How to choose E-23 help
- OSFI’s AI safety code: what we know so far
Sources
This template follows OSFI’s published guideline. The examples are made up. It isn’t legal or regulatory advice.