How to run a two-week practicum for KYC analysts

Give each KYC analyst ten working days on the live onboarding queue, deciding real held cases while a named assessor checks every decision against the record. Set the bar before day one: at least 95 in every 100 decisions correct and no missed red line, such as clearing a sanctions match. Sign off only on evidence the record holds.

What is a KYC practicum, and who owns it?

A practicum is supervised practice on real work. For a KYC analyst, the work is the cases an onboarding agent holds for a person: identity documents that do not match, screening hits, unclear ownership structures and source-of-funds explanations that need judgement. The analyst decides; the assessor checks; the record keeps both.

The head of onboarding or the MLRO's team owns the practicum, not the training function. They own the queue, the risk appetite and the consequences of a wrong clearance. Learning and development can schedule it, but the bar is set by the people who answer for the onboarding decisions. For the wider picture of why practical proof matters, see the guide to AI literacy and certification for people who work with agents.

How do you run it, step by step?

1. Person (onboarding lead): choose the queue. Pick a live queue where the agent works at depth 2 or 3: it drafts verification outcomes and requests documents, and holds anything outside its bounds. Avoid a queue so quiet the analyst sees fewer than a handful of holds a day.
2. Person: write the red lines down. Clearing a possible sanctions or PEP match, accepting a document the agent flagged as altered, and approving a corporate customer without identified beneficial owners. Missing any one ends the practicum regardless of the overall score.
3. Agent, depth 2: drafts the case summary. For each held case the agent sets out what it checked, which sources it used and which bound held it. The analyst reads this, then the evidence itself.
4. Person (analyst): decides every held case. Approve, edit or decline, with a written reason that cites the evidence. A decision without a reason counts as wrong.
5. Agent, depth 3: requests missing documents within its bounds. Routine follow-ups go out without the analyst; the analyst checks a sample to see the agent asked for the right thing.
6. Person (assessor): marks each decision the same day. Correct, incorrect or correct with a weak reason. Disagreements are discussed while the case is fresh.
7. Person (assessor): runs a midpoint review on day five. Patterns matter more than single errors: an analyst who clears declared source-of-funds statements without asking for evidence needs coaching now, not at the end.
8. Person (assessor and onboarding lead): sign off or extend. Sign-off rests on the record, not on the assessor's impression.

Which controls keep it honest, and what evidence do they leave?

• Same queue, same rules as production. The analyst works on real customers, so their decisions take effect. The evidence is the filed verdict for each case, with the analyst's name on it.
• Red-line cases are seeded if the queue lacks them. If two weeks pass without a single screening hit, the assessor adds past cases replayed in a training workspace. The record notes which cases were replayed, so nobody mistakes them for live decisions.
• The assessor is independent of the analyst's line manager. The evidence is a named assessor on every mark.
• Provenance is checked, not assumed. An analyst should see that a passport scan is external evidence and a registry extract is signed. Treating a customer's declared ownership chart as if it were verified is a marked error.

What should you measure?

Measure four things, all straight from the record:

• Decision accuracy: correct decisions as a share of all decisions, against the 95 in 100 bar.
• Red-line misses: must be zero.
• Reason quality: how often the assessor marks a decision "correct, weak reason". A rising count means the analyst is guessing well rather than reasoning.
• Time to decide by hold reason: not as a pass criterion, but because a KYC analyst who takes ten minutes on a simple address mismatch and two on a PEP match has the effort the wrong way round.

Volume is not a measure of competence. An analyst who decides 200 cases badly has not done better than one who decides 80 well.

Where does a KYC practicum usually go wrong?

The queue is too clean. Onboarding teams sometimes route the hard cases to seniors during the practicum to protect the customer. The analyst then passes on easy work and meets their first ownership chain alone. Route the hard cases to the analyst and have the assessor check them before they take effect.

The assessor rubber-stamps. An assessor who marks fifty decisions a day without opening the evidence is producing a score, not an assessment. Limit the number of analysts per assessor.

Nobody connects the result to approval rights. If an analyst who failed the practicum can still approve held cases the next morning, the practicum was ceremony. Compare how other teams handle this in the playbooks for credit analysts and fraud analysts, and use the practicum log template for KYC analysts to keep the marks in one place.

Where the Academy fits

The practicum is part of Certified Pryme Operator, the Academy's certificate for people who run a live workspace. It is two weeks of real review decisions, at least 95 in every 100 correct and no missed red line, signed off by a Pryme assessor, alongside a written exam of 40 questions where 32 correct is a pass. The certificate is valid for 12 months and can be checked at source through VerifyDoc.ai.