AML Policies and Procedures: What to Write Down
Canadian reporting entities must keep written, approved, risk-based policies and procedures as part of the compliance program required by PCMLTFA s. 9.6 and PCMLTFR ss. 156–157. This article separates policies (what and why) from procedures (how), then shows how version control and named control owners turn the documents into a program you can operate and evidence.
Reader question
What do we actually need to write down in our AML policies and procedures, and how do we keep those documents current and accountable?
The written program the law expects
PCMLTFA s. 9.6 requires every reporting entity to establish a compliance program that is reasonably designed, risk-based, and effective. PCMLTFR ss. 156 and 157 set out the elements that program must contain: an appointed compliance officer, written policies and procedures, a documented risk assessment, ongoing training, and a review of the program's effectiveness every two years.
The policies-and-procedures element is where most of the practical writing happens. FINTRAC's compliance program guidance expects those documents to be written down, kept current, and formally approved — the current guidance describes the approval expectations in detail. A program that exists only in the compliance officer's head, however well it runs day to day, does not meet the written standard the regulations describe.
Policies say what; procedures say how
A policy states what the business must do and why: "we identify every client before providing a service, because the law requires it and our risk assessment rates onboarding as a key control point." A procedure states how staff actually do it: which verification method, which system, which fields to check, what to do when a check fails, and who to escalate to.
The most common documentation gap is a well-written high-level policy with nothing underneath it — no operating procedure for onboarding, monitoring, or reporting that a new hire could actually follow. A payroll platform that briefly holds employer funds may have a policy committing it to monitor transactions, but if no document says which reports get pulled, how often, and what an analyst does with a hit, the commitment has no operating substance. A useful test: hand the procedure to someone who has never done the task. If they cannot complete it without asking questions, the procedure is a policy in disguise.
Version control you can run in a spreadsheet
The question an examiner or auditor eventually asks is not "do you have a policy?" but "which version of the policy applied on this date, and what did it say?" Many businesses update their documents diligently and still cannot answer, because each update overwrote the last.
A simple system closes the gap: give every document a version number and an effective date; keep a one-page change register recording what changed, who approved it, and when; and retain every prior approved version instead of overwriting the file. None of this requires special software — a naming convention, a register, and an archive folder are enough, provided they are used every single time a document changes.
Control ownership: every requirement gets a name
A policy without an owner is hard to operate and hard to evidence. Control ownership means every requirement in the program carries four things: a named owner (a person or role, not a department), an evidence standard (the artifact that shows the control ran), a frequency (how often it runs or is checked), and an escalation path (what happens when it fails or the owner is unavailable).
Map owners across the whole program — KYC, transaction monitoring, reporting, training, and records — and confirm each owner actually knows what they own. An owner who first learns of the assignment during an examination is, in practice, no owner at all. A one-page control register listing each requirement with its owner, evidence, frequency, and escalation route turns the policy binder into something the business can operate and demonstrate.
Keep the paper aligned with what actually happens
Documents drift. The business launches a product, changes a vendor, or quietly adjusts a workflow, and the written procedure ends up describing a process nobody follows. Drift is worse than a known gap, because it means the documented program and the operated program are two different things.
The two-year effectiveness review required by the regulations is the formal backstop that tests whether the documents still match operations, but well-run programs do not wait for it. They treat product launches, new customer segments, and regulatory changes as triggers to re-read the affected procedures and issue a versioned change where needed. When a review does find drift, the fix is recorded the same way as any other change: what changed, who approved it, and when it took effect.
At a glance
- PCMLTFA s. 9.6 and PCMLTFR ss. 156–157 require a written, risk-based compliance program: a compliance officer, policies and procedures, a risk assessment, training, and a two-year effectiveness review.
- Policies state what the business must do and why; procedures state how staff actually do it — a policy with no operating procedures underneath is the most common gap.
- Version control means you can show which version of a document applied on any given date: version numbers, a change register (what changed, who approved, when), and retained prior versions.
- Every requirement needs a named control owner plus an evidence standard, a frequency, and an escalation path — across KYC, monitoring, reporting, training, and records.
- Documents must match actual operations; treat product launches and process changes as triggers to update, and let the two-year effectiveness review act as the backstop.
Common mistakes
- Writing a high-level policy with no operating procedures a new hire could follow for onboarding, monitoring, or reporting.
- Overwriting documents on each update, leaving no way to show which version was live on a given date.
- Keeping no record of what changed, who approved it, and when.
- Publishing policies with no named owner — or naming owners who do not know they own the control.
- Assigning an owner but no evidence standard, frequency, or escalation path, so the control cannot be demonstrated.
- Letting documented procedures drift out of line with what the business actually does, then discovering the gap during a review.
Sources
Regulatory anchor: PCMLTFA s. 9.6; PCMLTFR ss. 156 and 157.
This content is general education and industry perspective. It is not legal advice, does not create a solicitor-client relationship, and does not replace the PCMLTFA, the PCMLTFR, FINTRAC guidance, or advice from qualified legal counsel. It does not guarantee regulatory or bank acceptance. Confirm current law, current FINTRAC guidance, and the full facts before relying on it for a business decision.