How to run a two-week practicum for support team leads

Support team leads are tested on two kinds of decision: the escalations the agent holds for them, and the corrections they make to how the agent and team work. For two weeks an assessor marks both. The bar is 95 correct decisions in every 100 and no missed red line, such as a formal complaint closed as an ordinary ticket.

Why is a team lead's practicum not just a harder reviewer practicum?

A customer support agent reviews tickets. A team lead does that for the hardest tickets and also decides what happens next: whether a pattern of holds means the agent's brief needs changing, whether a reviewer on the team is approving too freely, whether a refund limit is set in the wrong place. Those decisions shape hundreds of future tickets.

The head of support owns the practicum; the assessor is an experienced lead from another team. For the wider certification picture, see the guide to AI literacy and certification for people who work with agents. The lead's own team members follow the customer support agents' practicum, and it helps if leads have seen that one from the inside.

What are the steps?

1. Person (head of support): give the lead the escalation queue for their team. Refunds above limit, account changes, complaints and anything a team member escalated.
2. Person (head of support with complaints owner): set the red lines. Approving a refund or credit above the lead's authority; treating a complaint that meets the formal definition as a routine ticket; making an account change without the identity checks policy requires; approving a backlog of holds in bulk without opening them.
3. Agent, depth 3: resolves routine tickets within bounds. Order status, password resets through the approved flow, refunds under the limit.
4. Agent, depth 2: drafts replies and actions for held tickets. With the reason it held and the sources it used.
5. Person (lead): decides every escalated ticket. Approve, edit, decline or route, with a reason.
6. Person (lead): proposes at least two corrections to the agent's brief or sources. Based on patterns in the holds. The correction goes through the normal change route; it is not made directly.
7. Person (lead): reviews a sample of one team member's decisions. And records coaching notes.
8. Person (assessor): marks every decision and correction. Sign-off or extension on day ten.

Which controls matter, and what evidence do they leave?

Bulk approval is visible. The record shows when a lead approved many holds in a short window. The assessor opens a sample of them; any that were wrong count individually, and the bulk approval itself is a red-line miss if items were not opened.

Complaints follow their own route. Where a complaint has a formal definition, through regulation or the organisation's own policy, the agent flags tickets that meet it. The lead's decision to log or not log is recorded with a reason.

Brief corrections are evidenced. A correction is proposed with the tickets that show the pattern. Qualification, not training: nothing changes in the model; what changes is the agent's brief, sources or bounds, and the change is recorded.

Coaching notes are specific. "Be more careful" is not a coaching note. "Check the order date before approving a late-delivery refund" is.

What should be measured?

• Accuracy against the 95 in 100 bar, across escalations and corrections.
• Red-line misses: zero.
• Correction quality: whether proposed brief changes, once made, reduce the hold reason they targeted.
• Complaint capture: every ticket meeting the definition logged as one.
• Coaching specificity: assessed by the assessor reading the notes.

The practicum log template for support team leads has separate sections for escalations, corrections and coaching.

Where does a lead's practicum go wrong?

The lead clears the queue and does nothing else. Busy leads become super-reviewers. If the fortnight ends without a single brief correction or coaching note, the practicum has not tested the role.

Corrections are made directly. A lead who edits the agent's refund wording on the fly has changed the agent's behaviour without a record. The correction must go through the change route.

Complaints hide in tone. Customers rarely use the word "complaint". The assessor checks whether the lead caught expressions of dissatisfaction that meet the definition.

The IT administrators' practicum shows a similar mix of routine changes and privileged requests. Before sign-off, settle with the lead whether support team leads need a certificate to approve AI agent actions in your organisation.

Where the Academy fits

Team leads are among the people the Operator track is written for. Its modules on human review, teaching the agent and running the rhythm match the three parts of this practicum, and the practicum itself is two weeks of real decisions, at least 95 in every 100 correct and no missed red line, signed off by a Pryme assessor. The resulting certificate is valid for 12 months.