What shapes an AI project in Canada
The rules that govern an AI agent in Canada are mostly rules that existed before AI. They cover personal information, messages, calls and language, and an agent has to follow them as an employee would.
| Topic | What it means for your agent | Decide early |
|---|---|---|
| AI legislation | A federal AI bill lapsed in 2025, and the privacy bill tabled in 2026 has no AI chapter. For now, privacy law, human rights law and regulators' guidance apply to automated systems | Who in your company owns the decision to automate a task |
| Personal information | PIPEDA and the provincial laws expect a clear purpose and meaningful consent for what you collect. Using customer records to answer that customer is one purpose. Using them to train a model is another | Which data the agent may read, and what is kept out of prompts |
| Automated decisions in Quebec | When a decision rests on automated processing alone, Quebec's law says the person is told, can ask what information and reasons were used, and can have a human review it | Which steps stay with a person, and how each automated decision is logged |
| Messages and calls | CASL sets consent, identification and unsubscribe rules for commercial emails and texts. The telecom regulator restricts automated sales calls and runs the National Do Not Call List | What the agent may send unprompted, and to whom |
| Language | Customers in Quebec expect service in French, and French in Canada has its own vocabulary. An agent is tested in both languages with your real questions | Who reviews French answers before launch |
| Where data is processed | Several model and cloud providers can process requests in Canadian regions. That matters when a contract, or a Quebec assessment, limits where information may go | Which provider and region the agent uses, and why |
Agent workflows we build for Canadian businesses
Each of these keeps a person in charge of the step that carries risk. Our catalogue of AI agent solutions describes many more.
Bilingual customer support
The agent answers from your own documents, in the language the customer wrote in.
- Question arrives in English or French
- Agent finds the answer in your content
- Replies in the same language
- Hands over to staff when unsure
- Conversation saved to the customer record
Invoice and document intake
For finance teams that retype supplier invoices into an accounting system.
- Document received by email
- Fields read and checked
- GST and provincial tax lines matched
- Exceptions sent to a person
- Approved entry posted to accounting
Booking and reminders with consent
For clinics, service companies and anyone whose calendar is their revenue.
- Customer asks for a time
- Agent offers open slots
- Consent for reminders recorded
- Reminder sent by text or email
- Reschedule handled in the same thread
Internal knowledge assistant
Staff ask about policies, products or procedures and get an answer with its source.
- Employee asks a question
- Agent searches approved documents
- Answer shown with the source
- Access limited by role
- Gaps reported to the content owner
What we build
Customer-facing agents
Chat, email and messaging agents grounded in your content. See the customer support agent for a worked example.
Back-office automation
Agents and rule-based workflows that read documents, update records and route exceptions. Our comparison of AI agents and workflow automation helps you pick.
Bilingual behavior
Prompts, test sets and fallback messages written and reviewed in English and French, so quality is measured in each language separately.
Consent and logging
A record of what the agent was allowed to do, what it did and why, kept where your privacy officer can reach it.
Human review steps
Approval queues for the actions that matter, such as refunds, refusals and anything sent to a regulator.
Integration with your systems
Connections to your CRM, help desk, accounting and calendar tools, described in AI integration services.
What we need from you
- The task you want automated, and how it is done today
- Real examples of questions, documents or cases, with personal details removed
- Your privacy adviser's view on consent and on Quebec, if it applies
- A reviewer for French output
- Someone who will judge the agent's answers each week of the pilot
Ownership and licensing terms are defined in the project agreement.
A readiness checklist for an AI pilot in Canada
Before the pilot starts
0 of 8 checkedOur guide to planning an AI pilot with measurable success criteria turns this list into a plan.
What to do next
Describe the task you would like to hand to an agent and how it is handled now. We will tell you whether an agent, a simpler automation or neither is the right tool, and what a pilot would involve. Start a conversation or book a 30-minute call.
Common questions
Is there an AI law in Canada that we must follow?
No AI-specific federal statute was in force when this page was reviewed. A proposed act lapsed in 2025, and the 2026 privacy bill does not include one. That does not leave AI unregulated. Privacy law, human rights law, consumer protection and sector regulators all apply to what an automated system does. Quebec's privacy law also has a specific rule on decisions made by automated processing alone.
Can an AI agent serve customers in both English and French?
Yes, and for a business that serves Quebec it should. Current language models write well in both, but quality still has to be tested with your own questions, your product names and Canadian French usage. We build separate test sets for each language and ask a French speaker on your side to review answers before launch and at intervals afterward.
Do we have to tell people they are dealing with an AI?
Telling them is the safer course everywhere and a legal duty in specific cases. In Quebec, a person must be informed when a decision about them is made by automated processing alone. Federal privacy regulators have also asked organizations to be open about their use of generative AI. We design agents to say what they are at the start of a conversation.
Can customer data be sent to an AI model hosted outside Canada?
Often yes, with conditions. Federal law leaves you accountable for the information wherever it is processed. Quebec expects an assessment before personal information leaves the province. Contracts with public bodies or health organizations may set a location. Where location matters, we use a provider that processes in a Canadian region and keep personal details out of prompts when the task allows.
Where to go from here
Related serviceAI AgentsAgents that work inside your systems, with approval where it matters