Industries · Credit unions
AI for Canadian credit unions
Member letters, lending policy, board packages, regulator questions — the paperwork a credit union runs on, with answers that cite the policy they came from.
01
What it does for a credit union
- 01
Policy and compliance Q&A
the lending policy, the deposit policy, the privacy code, the board’s governance manual: ask in plain language, get the clause.
- 02
Member-service drafting
rate-change notices, hardship responses, complaint acknowledgements, branch announcements, in the tone the credit union uses.
- 03
Staff enablement
a new member-service rep asks the questions a mentor would otherwise answer five times a day, and gets answers grounded in the current policy rather than last year’s PDF.
- 04
Board and committee prep
a plain-language summary of a long policy document, with the sections that changed called out.
- 05
Regulator and audit questions
turn an examiner’s information request into a short index of where each answer lives in your own documents.
02
OSFI E-23 and provincial regulators
Federally regulated institutions work under OSFI’s Guideline E-23 on model risk management, which asks an institution to inventory its models, understand their limitations and govern their use. Provincially regulated credit unions answer to their own provincial regulator and to provincial privacy law.
Digby is not a credit-scoring or decisioning model, and it should not be used to make lending decisions. Where it helps is the documentation around those models: explaining a policy, drafting a member communication, summarising a guideline, and keeping the source passage attached to the answer.
We do not claim that using Digby satisfies E-23 or any regulator’s expectations. It gives you the artifacts — the cited source, the recorded policy version, the deletion controls — that make your own assessment easier to evidence.
03
Controls a risk committee will ask about
Accounts, conversations, memory and files are stored in Canada (Canada Central, Montréal) and processed by models served on Canadian-located GPU hardware. Your content is never used to train or fine-tune AI models. Memory and conversations are visible and deletable. Share links are opt-in and revocable, and nothing is visible outside a project unless someone creates one.
04
Keeping member information where it belongs
Most credit unions land in the same place: policies, procedures, product documentation and internal templates go into a project; identified member records stay in the core system. That keeps the useful part of the tool — the institutional knowledge — while leaving the member file where the record of record lives.
Where a project does hold sensitive material, roles keep it narrow: viewers read, editors work, admins manage, and only the people invited see the project at all.
05
Self-hosted when the requirement is absolute
Some institutions have a hard requirement that no document and no prompt leaves infrastructure they operate. Self-hosted Digby runs the full stack — database, interface, integrations and coding workspace — on your own servers with your own OpenAI-compatible model endpoints.
The same product, the same interface, the same projects; only the operator changes. It is a real option, not a press release.
06
A sensible first month
- 01
One policy area
pick the one staff ask about most and load it, with its current version date in the document name.
- 02
One drafting workflow
member letters or branch announcements, reviewed by the person who signs them today.
- 03
One rule written down
what may go in, what never does, and who reviews an output before a member sees it.
- 04
One review
after a month, read a sample of answers against the policy and see whether the cited passages held up.
Frequently asked questions
Does Digby make lending or credit decisions?
No. Digby is not a credit-scoring or decisioning system, and it should not be used to approve, decline or price credit. It works on policy Q&A, drafting and summarisation.
Is Digby OSFI E-23 compliant?
We do not claim compliance with E-23 or any regulator’s expectations. E-23 asks institutions to govern model risk themselves; Digby can support the documentation side by keeping cited sources, policy versions and deletion controls, and your own assessment remains yours to make.
Where is member information stored?
Anything you put into Digby is stored in Canada — the Canada Central region in Montréal — and processed by models on Canadian-located GPU hardware. Many credit unions keep member records in the core system and use Digby for policies and drafting instead.
Can we delete something we uploaded by mistake?
Yes. Files, conversations and individual memory facts can be removed, and share links can be revoked.
Can we run it on our own infrastructure?
Yes. Self-hosted Digby runs on your servers with your own model endpoints, so documents and inference stay inside the environment you operate.
Does Digby support French service?
Yes. Ask in French and the answer comes back in French, and French documents are read as readily as English ones — useful for member communications in bilingual regions.
Try asking Digby
Keep reading
Try Digby now.
No account needed. Models and data stay in Canada.