AI for Canadian public sector teams

Those four answers are the same whether the buyer is a city of forty thousand or a federal agency, and they are on this page in language you can paste into a briefing note.

1. 01

   Where is the data stored?

   In Canada, in the Canada Central region (Montréal): accounts, conversations, memory and uploaded files.

2. 02

   Where is it processed?

   The models that write answers are served on Canadian-located GPU hardware.

3. 03

   Can we delete it?

   Conversations, memory facts, files and whole accounts can be deleted; share links are revocable.

4. 04

   Who else sees it?

   Only the people invited to a project. Nothing is shared unless a link is created, and nothing is used to train AI models.

1. 01

   Policy and bylaw Q&A

   the procedural bylaw, the records-retention schedule, the procurement policy, the council handbook, each answered with the passage.

2. 02

   Plain-language rewriting

   a staff report or technical memo turned into something a resident can read, without changing what it says.

3. 03

   Meeting and agenda support

   draft summaries, talking points and public-facing explainers from documents the office already produces.

4. 04

   Bilingual service

   a document in English and a first draft in French (or the reverse) for staff to refine, never as a substitute for translation review.

5. 05

   Onboarding

   a new employee asks the questions a senior colleague would otherwise answer for a month.

Public bodies hold records that may be subject to access-to-information legislation, and to privacy statutes such as British Columbia’s FIPPA, Alberta’s FOIP, Ontario’s MFIPPA and Quebec’s Law 25. Those laws govern collection, use, disclosure and retention.

Digby stores data in Canada and gives you the deletion and access controls those regimes care about, but it does not make a public body compliant and we do not claim that it does. A records officer still decides what may be recorded, how long it is kept and what may be disclosed.

The practical implication for a records manager: treat a Digby project like any other shared drive holding working material, and apply your existing retention schedule to it rather than inventing a new one.

Agencies with a hard on-premises requirement can self-host: the full Digby stack on your own servers, pointed at your own model endpoints. Documents and inference stay inside the environment your IT team operates, and the product is the same one described on this page.

That is often the cleanest answer for a public body whose security review cannot approve a hosted service, and the most honest one for a body that simply prefers to own the whole stack.

1. 01

   Start on public documents

   bylaws, policies and published plans, where nobody has to make a judgment call about sensitivity.

2. 02

   Pick one workflow

   a weekly plain-language summary of a council agenda is a good first test.

3. 03

   Confirm your own rules first

   ask the records or privacy officer what may go in, before staff start uploading.

4. 04

   Name a reviewer

   anything that leaves the building gets read by a person who signs off on it.