Onboard a new support team lead in two weeks by teaching three things in order: how the agents decide, how the team reviews them, and what happens on the day something stops. Week one is observation and verdict reading. Week two is running the held queue, coaching one reviewer, and owning the lapse calendar under the outgoing lead's eye.
What does a support team lead own once agents are working?
The lead no longer owns most first replies. Agents draft them at depth 2, and in many teams they execute routine replacements at depth 3, and refunds at depth 4, within the business's limits. What the lead owns is the system around that: who reviews which held items, whether reviewers are making good calls, whether limits are set right, and how the team copes when an agent's certificate lapses on a Monday morning.
The support manager or head of customer operations owns the onboarding. The outgoing lead, or a peer lead, is the guide. How the role's day changes is covered in what changes in a support team lead's day when AI agents arrive.
Ten working days, step by step
- Day 1. Person (support manager): grants the lead read access and adds them to escalations. No approval or limit-change rights yet.
- Days 1 and 2. Person (new lead): maps the queue. Which agent handles which contact reason, at what depth, against which limits. A refund agent may execute refunds up to 100 at depth 4; above that it may only draft, and a person makes the refund.
- Days 2 and 3. Agent, depth 2 and 3: works the queue; the new lead sits with two reviewers. One experienced, one newer. Watching the gap between them shows the lead what coaching is needed.
- Day 4. Person (new lead): reads twenty verdicts and names the deciding bound. A refund of 140 refused by the business's own limit, so a person makes it. A request to change a customer's email address refused because identity changes sit outside the agent's role scope. A request to reverse a settled payment withheld at the platform floor, for every agent, at every depth.
- Day 5. Person (new lead): builds the lapse calendar. Every certificate lasts 90 days; a lapse, or a run that clears no depth, drops the agent to depth 1 the same day. The lead lists each date and decides who covers the extra held volume.
- Days 6 to 8. Agent, depth 2 and 3; person (new lead): runs the held queue allocation and approves escalations, with the outgoing lead countersigning limit-related calls.
- Day 9. Person (new lead): coaches one reviewer using that reviewer's own decisions from the record.
- Day 10. Person (support manager): reviews the lead's allocation, escalation and coaching notes, then hands over queue ownership and the lapse calendar. Changes to limits stay with the manager until the lead is certified.
| Step | The agent does | Depth | Stays with a person |
|---|---|---|---|
| Access and escalations | Runs as normal | As certified | Access grant |
| Map the queue | Answers by contact reason | 2 to 3 | Building the map |
| Sit with reviewers | Drafts and executes routine work | 2 to 3 | Every hold decision |
| Read verdicts | Files verdict and deciding bound | 2 to 3 | Explaining the bound |
| Lapse calendar | Drops to depth 1 on a lapse, or a run that clears no depth | 1 if so | Cover planning |
| Run the held queue | Drafts above limit for a person | 2 to 3 | Allocation and escalation calls |
| Coach a reviewer | No change | As certified | Feedback from the record |
Which controls matter in the handover, and what do they leave?
Separation of limit changes from queue running. The new lead runs the queue but cannot raise the refund limit. Evidence: the record shows any limit change and who made it.
Countersigned escalation calls in week two. Two names on every decision that touched a limit.
A written lapse calendar. Dates, owners and cover. If an agent drops to depth 1 and the queue floods, the calendar shows whether the team planned for it.
Coaching notes tied to real decisions. Not "be more careful" but "on these three refunds you approved a replacement where the order was already marked delivered by the courier, which is a signed fact".
What should the support manager measure?
- Reviewer accuracy before and after the lead's coaching, measured on decisions, not speed.
- Held items waiting past the team's own response target, and why.
- Escalations the lead sent upwards that the manager would have resolved, and the reverse.
- Whether the lapse calendar was right: did cover arrive when a certificate lapsed?
Where do new support leads trip up?
Managing handle time instead of decisions. Leads from a pre-agent team often push for faster held-queue clearing. Faster approval of held items means less scrutiny of exactly the items the agent was unsure about.
Treating agents as staff to performance-manage. An agent that keeps holding a type of query is usually missing a source or a bound, not working too slowly. The fix is to change its brief or sources, not to complain about it in a team meeting.
Not knowing who governs the agents. Some businesses have a named governance lead who owns limits and certification decisions; the lead should know who that is on day one. See what an AI governance lead is and what an AI governance lead does.
Frontline onboarding is different: compare onboarding a new customer support agent. The full approach is in training teams for AI agents.
Where the Academy fits
For teams, Pryme Intelligence Academy gives an admin view of progress, results, certificates and expiry dates. A support lead can use it to see which reviewers hold a current Certified Pryme Operator certificate and when each one is due for renewal, which sits next to the agents' own lapse calendar.
Questions people ask
Should a new support lead also review held items personally?
Yes, for some of the queue. A lead who never approves anything loses the feel for what the agents are holding and why. Most leads keep a small share of the queue for themselves.
Who decides the refund limit the agent works within?
The business, through whoever owns its own limits, often the support manager or a governance lead. The limit is one of the five bounds every action must meet.
What if the outgoing lead has already left?
Use the record. Every verdict and every human decision is filed, so the new lead can reconstruct how the queue was run. Pair them with a peer lead for the countersigning step.