Sidekick Orchestration

Guide for Canadian advisors

How advisors are using AI, and where the lines are.

Start with one question for compliance.

“Which AI tools are approved here, and can client information go into any of them?”

  • Chase Bernier · checked November 12, 2026
  • 15 sources
  • Not legal advice

AI can draft and summarize. You still decide, review and keep the record.

  • Usually fine

    In a firm-approved business tool, with a person reviewing.

    • Drafting emails, agendas and follow-ups for you to edit
    • Summarizing research, market commentary and long documents
    • Turning your own notes into tasks and reminders
    • First drafts of marketing, which still goes through your normal review
  • Ask your firm first

    Your firm, MGA or carrier may be stricter than the law. Solo? You are your own compliance team.

    • Meeting notetakers: client consent, where recordings live, and how they're kept as records
    • Putting any client personal information into an AI tool
    • Connecting AI to your email, CRM or client files
    • Whether and how you tell clients you use AI
  • Don't

    Most of these come from securities regulators. The last one is common firm policy.

    • Let AI make the suitability decision or the recommendation
    • Send anything to a client that no person has read
    • Talk to clients through an AI tool your firm can't capture and keep
    • Claim more about your AI use than is true ("AI washing")
    • Put client information in a personal or free AI account, or SINs, ID, account or card numbers into any AI

Sort these by your role and province in the advisor AI use map

Five checks before client information goes into AI.

  • It's a firm-managed business account.

    A firm-managed plan with training on your data turned off and an admin controlling access. Free and personal accounts are where client data leaks.
  • You know where the data is stored and for how long.

    Ask the vendor for data location, retention and certifications such as SOC 2. Quebec requires a privacy assessment and a written agreement before personal information leaves the province, including to an AI vendor.
  • It reaches only what it needs.

    Connect AI to the mailbox, folder or CRM fields the job needs. Let it read and draft. A person sends.
  • What it produces is kept.

    Transcripts, notes and AI drafts sent to clients can be business records. They should land in your firm's records system as well as the vendor's.
  • You know what happens if something leaks.

    Name who gets told and how fast. Privacy law can require you to report a breach. More on setup: the three settings that keep client data safe.

What regulators have actually said.

“Market participants are responsible for the outputs of the technology they use, including AI systems.”
Canadian Securities Administrators, Staff Notice 11-348, December 2024.
SourceWhat it saysWhat it means for you
CSA Staff Notice 11-348, Dec 2024“registrants cannot outsource registerable activity (e.g., trade suitability determinations for clients).”AI can help prepare. The suitability call stays yours.
CISRO guidance for insurance intermediaries, Nov 2024“Systematically check the content generated in this way before communicating it to another person.”Insurance advisors: check AI output before a client sees it.
AI washing, CIRO, records and breaches: 6 more lines
CSA Staff Notice 11-348Warns against “inaccurate, false, misleading, or embellished claims”, which it calls “AI washing”.Describe your AI use plainly in marketing.
CIRO Compliance Report 2026“we will be enquiring about the use of AI in dealers’ operations” (financial and operations exams)Expect AI questions in CIRO’s financial and operations exams.
CIRO Compliance Report 2026Looks for “client communications through non-approved channels”.Treat an AI chat or notetaker that touches clients as a channel that needs approval and capture.
CIRO Guidance Note 11-0349Retention applies “regardless of the method by which communication takes place” (written about social media, for investment dealers).Client messages are kept however they’re written or sent.
NI 31-103, s. 11.5Firms must keep records to “accurately record its business activities, financial affairs, and client transactions”.Keep the AI outputs you act on.
PIPEDA, s. 10.1Report to the federal Privacy Commissioner a breach that “creates a real risk of significant harm”.A leak through an AI tool is a breach like any other.
Four things people often get wrong

Recording calls: one-party consent is only half of it.

Criminal Code s. 184 allows recording with one party's consent. That only settles the criminal question. Privacy law still expects clients to know you're recording and why, and you stay responsible for any vendor that processes it.

BC has no mandatory breach reporting for private firms.

BC's private-sector privacy law still has no mandatory breach-reporting duty. BC's privacy commissioner “strongly recommend[s]” reporting anyway (OIPC BC). Federal law, PIPEDA, can still apply.

Insurance advisors: CISRO's guidance applies, plus fair treatment.

Canada's insurance licensing regulators (CISRO) published guidance for intermediaries in November 2024: “Systematically check the content generated in this way before communicating it to another person” and “Don't share confidential information in public and open solutions” (CISRO). Fair-treatment rules also apply. Quebec's AMF AI guideline and OSFI's model-risk guideline E-23 (federally regulated banks and insurers) both take effect May 1, 2027. They apply to insurers and other financial institutions, not directly to individual advisors (OSFI E-23).

What's coming.

No federal AI law exists today. Bill C-36, a federal privacy bill, was introduced June 15, 2026 and is awaiting second-reading debate as of November 12, 2026 (LEGISinfo).

Sources (15 links)

How advisors are using AI.

  • 65%of Canadian planners say their firm uses AI, up from 36% in 2025FPSB 2026
  • 37%say their firm has a comprehensive AI policy, up from 16%FPSB 2026
  • 53%worry about the accuracy of AI output, up from 38%; 43% worry about privacyFPSB 2026

Worldwide, email and notes lead (61%, Natixis). About 43% of practices use a notetaker such as Jump or Zocks (T3, US). Only 18% of US advisors rate AI highly reliable for investment advice (Morningstar). See one advisor's pre-call briefs.

A 30-day path: one tool, one job, rules in writing.

  1. Week 1: Get one tool approved. Ask compliance which tools are approved, or whether Microsoft 365 Copilot can be turned on. Solo? Ask your MGA or carrier.
  2. Week 2: Write a one-page AI note for your practice. What AI does, what never goes in, who reviews. Start from the one-page AI policy template.
  3. Week 3: Start one low-risk job. Follow-up emails, research summaries or annual review prep. Note the time before and after.
  4. Week 4: Decide on meeting notes. Ask compliance for consent wording, storage and archiving. Already using one? Check those now.

Then add one job at a time. Next: a sample AI setup record or how to adopt AI at your company.

Not sure which job to start with? The free AI Opportunity Map suggests a first project and how to handle your data. About 3 minutes.

Outside Canada? Research your own rules.

Copy this prompt into an AI tool with web search turned on, fill in the brackets, and run it. You get a sourced brief for your compliance lead: primary sources only, a quote and date for every rule, a skeptical self-check, and a Confirmed, Likely or Gap label on each finding. Use a business account, and never paste in client information.

Research prompt · works in ChatGPT, Claude, Copilot or Gemini

Tested November 12, 2026 for a UK adviser and an Ontario insurance agent: both runs quoted dated sources, flagged gaps and dropped what they couldn't confirm. The Ontario run found CISRO's 2024 insurance guidance, now in this guide.

Read the full prompt first
I'm preparing for a conversation with my compliance lead about using AI in my advice or insurance practice. Research only. This is not legal advice, and I'll confirm everything with my compliance team.

About me (fill in):
- Country and province or state: [e.g. United Kingdom / Ontario, Canada / Texas, USA]
- My role, licence and type of firm: [e.g. FCA-authorised adviser at a 12-person wealth firm / mutual fund dealer rep / solo life insurance agent with an MGA]
- Who regulates me, if I know: [e.g. FCA / CIRO / FSRA / state insurance department]
- What I want to use AI for: [e.g. a meeting notetaker, drafting client emails, summarizing research]
- Tools I'm considering: [e.g. Microsoft 365 Copilot, ChatGPT Business, an advisor notetaker]

Rules for you:
1. Use web search. If you can't search the web, stop and tell me. Don't answer from memory.
2. Use primary sources only: regulator websites, legislation and official guidance. Use news or law-firm articles only to find those sources, never as the source.
3. For each requirement, give the rule in one plain sentence, a short exact quote, and the document name, issuer, date and link. If a source won't load, say so.
4. Cover what applies to my use: advice and suitability duties; record-keeping and capturing client communications; privacy and data protection, including where data may be stored and breach reporting; recording calls and client consent; marketing claims about AI; any AI-specific guidance; and anything coming into force in the next 12 months.
5. Then check your findings as a skeptical compliance officer would. Reopen each source, confirm every quote and date, look for newer guidance that replaces it, and remove anything you can't confirm. Tell me what you changed.
6. Mark each finding Confirmed (quote checked at the source), Likely (source found, wording not confirmed) or Gap (no clear source found).
7. Tell me which answers depend on details about my firm that I didn't give you.
8. Don't ask for or include any client information.

Give me back, in under 600 words:
- The 3 to 5 things that matter most for my use, in plain language.
- A table: requirement | what it means for me | source (quote, date, link) | status.
- Gaps: what you couldn't confirm, and who to ask.
- 3 specific questions to take to my compliance lead.
- The date you checked.
Help improve this guide

What's wrong, missing or different at your firm?

The email opens with three short questions: what's wrong, where you want a clearer line, and which tools your firm allows. Answers stay anonymous unless you say otherwise.

Does ChatGPT recommend your practice? AI Search Check, about 2 minutes
Want to talk it through? Book a free 30‑minute call