Canadian payment and money services firms should train AML investigators to check the evidence an agent compiles, review its draft suspicious transaction reports and transfer reports, confirm transfer information is complete, and keep every decision to report, restrict or exit with a named person. Canada's AML law already requires an ongoing training programme; agents change what it must cover.
What does an AML investigator at a Canadian payments firm do?
Many Canadian payments firms are registered money services businesses: remittance providers sending funds to families abroad, foreign exchange dealers, payment platforms and card programme managers. Their investigators work alerts from transaction monitoring, examine customers whose activity no longer matches their profile, and decide whether there are reasonable grounds to suspect money laundering or terrorist financing.
The job is also heavy on prescribed reporting. Large international electronic funds transfers must be reported, as must large cash and virtual currency transactions where they apply, alongside suspicious transaction reports. Investigators check that transfer information travels with each payment as the international standards require, handle sanctions screening hits, and answer production orders. Much of the day is document assembly rather than analysis.
Where do agents help, and at what depth?
Alert enrichment (depth 1). The agent gathers the customer's identification records, transaction history, linked accounts, counterparties and prior alerts, each fact marked with its source.
Transfer information checks (depth 1). The agent checks that originator and beneficiary information is present and consistent on incoming and outgoing transfers and flags gaps.
Prescribed report preparation (depth 2). For threshold-based reports such as large international transfers, the agent prepares the report from sealed transaction data for an investigator to review and submit.
Suspicious transaction report drafting (depth 2). Once the investigator has formed suspicion, the agent drafts the narrative in the structure the firm uses. The investigator edits it and decides to file.
Screening hit triage (depth 2). The agent compares a hit's details with the customer record and drafts a disposition for a person to confirm.
Depth here is earned on the firm's own cases. A firm might certify report preparation quickly, because the data is sealed and the rules are prescriptive, while narrative drafting stays under full review for longer.
How do investigators review the agent's decisions?
Every verdict is filed with the bound that decided it and the sources used. Investigators should check provenance first: transaction data from the ledger is sealed, an identity document verified through a reliable source is signed, a customer's account of the source of funds is declared, and a news item is external. A narrative that treats the declared story as established is a narrative to rewrite. How AML investigators review an agent's decisions sets out the method.
During training, investigators should keep a log of their review decisions: what the agent proposed, what they decided, and why. The practicum log template for AML investigators gives the fields. The log doubles as evidence for the training programme and shows supervisors that review is real, not a click.
What stays with the investigator?
Forming reasonable grounds to suspect, deciding to file, deciding to restrict or exit, answering law enforcement, and every message to a customer whose activity is under review. These are judgements for which the firm and its compliance officer answer, and they are why investigators remain the real control in an agent deployment.
Investigators also need to know when the agent's authority has changed. An agent's certificate lapses after 90 days, and a run that clears no depth drops it to depth 1 the same day. Reports it used to prepare will then arrive as flags, which matters when a reporting deadline is close. For the longer view of how the role changes, see will AI agents replace AML investigators?.
Which Canadian rules shape oversight and training?
Canada's proceeds of crime and terrorist financing law requires reporting entities, including money services businesses, to maintain a compliance programme with a compliance officer, written policies, a risk assessment, an ongoing written training programme and a regular effectiveness review. Canada's financial intelligence unit, which also supervises compliance, receives the reports. When an agent changes how investigations are run, the training programme and policies should change with it.
OSFI supervises federally regulated financial institutions. Where a payments business is part of one, OSFI's Guideline E-23 on model risk management applies to models, including AI, used in its operations. The FCAC oversees consumer protection obligations of federally regulated financial institutions. PIPEDA governs the personal information the agent processes in the private sector. There is no federal AI act in force, so expectations about AI come through these existing rules.
The FATF Recommendations set the international standard for the risk-based approach, customer due diligence and transfer information that Canadian rules reflect.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
A 90-day plan for the investigations team
- Days 1 to 30. Investigators learn depth, verdicts and provenance using the firm's own closed files. The agent runs at depth 1 on enrichment and transfer information checks. The compliance officer updates the written training programme to include agent oversight.
- Days 31 to 60. The agent prepares threshold reports and drafts narratives at depth 2. Investigators review everything and keep a practicum log. Weekly sessions agree on disputed cases; corrections go into the agent's brief.
- Days 61 to 90. Investigators who approve the agent's held items complete certification and supervised practice. Report preparation is tested in Attest before any move beyond depth 2. The effectiveness review notes the change.
The AI agent training plan template for AML investigators gives owners and dates. For analysts, QA and the compliance officer, see AI agent skills by role.
Where the Academy fits
Pryme Intelligence Academy gives investigators a training record that fits a written programme: Foundations for everyone, then the Operator track for those who review agent work, with a 40-question exam (32 to pass) and a two-week practicum signed off by an assessor. The admin view shows each person's results and certificate expiry.
Questions people ask
Can the agent submit threshold reports automatically?
Not at first. After a sustained record of accurate preparation and certification, a firm may allow submission of routine threshold reports within strict bounds, with any anomaly held.
Does the training programme need to name the agent?
It should describe how agent outputs are reviewed and by whom. Supervisors look for training that matches how the work is actually done.
What should investigators do with an incomplete transfer?
The agent flags it; the investigator decides whether to request the missing information, hold the transfer or consider it in a suspicion assessment.