PrymeIntelligence

AI agent skills by role·Template

AI agent competency matrix for onboarding team

A competency matrix for customer onboarding teams using AI agents: the KYC skills, target levels and evidence for analysts, seniors and the manager.

Customer onboarding teams need a competency matrix built on one idea: an agent can collect and compare evidence about a new customer, but people decide who becomes one. KYC analysts must weigh each document by where it came from, judge screening hits the agent holds, and approve or refuse higher-risk customers with reasons a supervisor could follow.

What is the onboarding matrix for?

Onboarding is where customer due diligence happens, and it is a standard target for agents. An agent can read an application at depth 1, request missing documents at depth 2, and in some firms open low-risk accounts at depth 3 within tight bounds. Each step needs a person who understands what the agent relied on. The general meaning of onboarding is simple; the skill is in judging evidence.

The core of that skill is provenance. A passport scan uploaded by the applicant is declared. An identity check returned by a verified issuer is signed. A company record pulled from the firm's own systems of record is sealed. An analyst who treats these as equal will approve the wrong people. The matrix also fits within the broader map of AI agent skills by role.

How do you fill it in?

  • Put the onboarding steps down the side, in the order a case moves.
  • Put the roles across the top: KYC analyst, senior analyst, onboarding manager and agent owner.
  • In each cell, enter one of four levels: 0 no role, 1 understands, 2 decides in the live queue, 3 approves exceptions and coaches.
  • Add a column for the evidence that proves the level, and one for the red line at that step: the thing nobody approves, whatever the queue looks like.

The template

Onboarding step Agent's part Analyst Senior Manager Agent owner Evidence Red line
Application intake Reads and checks completeness
Identity evidence Compares documents and checks
Ownership and control Drafts structure summary
Screening Runs checks and drafts disposition
Risk rating Proposes a rating with reasons
Enhanced due diligence Drafts requests and summary
Decision Opens low-risk accounts within bounds, or holds
Customer messages Drafts requests and outcomes
Record Files every verdict

A worked example

A hypothetical business bank's onboarding team has twelve analysts, four seniors, a manager and an agent owner from operations.

Step Agent's part Analyst Senior Manager Owner Evidence Red line
Intake Completeness check, 1 2 2 1 1 Queue sample None
Identity Document comparison, 1 2 3 2 1 50 files re-checked Approve on declared evidence alone
Ownership Structure summary, 2 1 3 3 1 Case review Unresolved controller
Screening Draft disposition, 2 2 3 3 1 Calibration session Close a possible match without review
Risk rating Proposed rating, 2 2 3 3 2 Rating audit Lower a rating without a reason
EDD Drafts summary, 2 1 2 3 1 Manager sign-off Approve high risk below manager
Decision Opens low risk, 3 2 3 3 2 Daily sample of 10 Open outside bounds
Messages Drafts requests, 2 2 2 2 1 Message review Disclose screening reasons
Record Files verdicts 1 2 3 3 Walkthrough None

Analysts are at level 1 on ownership and control. Untangling who really controls a company is senior work, and the agent's structure summary is a starting point, not a conclusion. Seniors reach level 3 because they are the ones who must spot a nominee arrangement the summary missed.

The decision row reflects a common design: the agent opens low-risk personal or simple business accounts at depth 3, and an analyst samples ten a day. If the agent's certificate lapses after 90 days, those openings become held items, and the analyst queue grows sharply until it is recertified. The reviewer rota for the onboarding team should show who absorbs that.

Customer messages carry their own red line. A draft that tells a customer their application is delayed "because of a screening match" says too much, and the analyst who sends it answers for it. The support team's competency matrix covers the same messaging skill for after-sales contact.

Where the Academy fits

Pryme Intelligence Academy's Operator track trains onboarding analysts and seniors on the real product, with module tests set in their own industry and role. Its practicum marks two weeks of real review decisions and requires no missed red line, which matches the red-line column above. The Foundations badge is enough for the agent owner's level 1 cells.

Questions people ask

Should the agent ever approve a high-risk customer?

No. High-risk approvals should be held for a named person, usually the onboarding manager. The agent prepares the file; the person decides and signs.

How do you test an analyst's skill with provenance?

Give them a set of sample files where some evidence is declared, some signed and some sealed, and ask them to rate each file. Analysts who rate a file well on declared evidence alone need more practice.

Who should own the onboarding agent?

The onboarding manager or the operations lead who owns the process. IT builds and connects it, as the IT team's competency matrix describes, but ownership belongs with the business.

Train the people who run your agents.

Pryme Intelligence Academy: free courses on the real product, audio lessons, labs in a free workspace, and certificates anyone can check on VerifyDoc.ai. Teams get seats, an admin view and a free readiness assessment.