Operations managers run many agents across many teams, so their readiness is about the calendar as much as any one change. Stagger certificate dates so no two critical agents lapse in the same week, plan capacity for a failed run, and check every 90 days that each queue still has certified approvers. Then make sure each team lead briefs their own delta.
What does recertification mean for an operations manager?
A team lead sees one agent's cycle. An operations manager sees a dozen, each lapsing 90 days after its last certification. Each recertification can move the agent up or down a depth, and a run that clears no depth drops that agent to depth 1 the same day. Across a dozen agents, that means at least one change somewhere most weeks.
The manager's own tools are agents too. A reporting agent reads the workspace at depth 1 and compiles the weekly picture: certificates due, verdict counts, held items by age, refusals by deciding bound. A planning agent drafts the review rota and the capacity plan at depth 2. The manager decides the rota and signs off capacity.
Two clocks run at once. Agent certificates lapse every 90 days. People's professional certificates last 12 or 24 months. The manager has to watch both, because a queue can lose its last certified approver on the same day its agent drops a depth. A short account of the agent clock is in what recertification is.
Who owns what?
The operations manager owns the calendar, fallback capacity and approver coverage. Team leads own briefing their teams on each change. Builders own recertification. Risk owns the organisation's own limits. The manager is the person who notices when these are out of step.
The steps
- Agent, depth 1: build a 90-day view of every agent's expiry, every approver's certificate expiry, and each queue's depth in use.
- Person (operations manager): stagger. Move recertification dates so that critical agents, and agents sharing a queue, never lapse together, and none land on known peaks.
- Agent, depth 2: draft a fallback capacity plan for each critical agent: extra human hours needed if it drops to depth 1.
- Person (operations manager): approve the plan and agree it with team leads.
- Person (builders): recertify each agent in turn.
- Person (team leads): write and brief the change note for their team within a week of each result.
- Agent, depth 1: check approver coverage after each change: at least two certified approvers per depth 3 or 4 capability on every shift.
- Person (operations manager): run the weekly rhythm: held items by age, refusals by bound, and briefings outstanding.
| Step | The agent does | Depth | Stays with a person |
|---|---|---|---|
| 90-day view | Compiles expiries and depths | 1 | Spotting clashes |
| Staggering | Nothing | none | Moving dates |
| Fallback plan | Drafts capacity needs | 2 | Approving and funding |
| Approver coverage | Checks certified people per shift | 1 | Fixing gaps |
| Weekly rhythm | Compiles the picture | 1 | Decisions and follow-up |
| Briefings | Reminds team leads | 2 | Delivering them |
What should operations managers keep relearning?
Reading the workspace at scale. Held items by age, refusals by deciding bound and drafts returned with edits tell the manager where an agent and a team are out of step. The skill is spotting a trend across agents, not chasing one item.
Usage and cost after depth changes. When an agent drops from depth 3 to depth 2, people do more of the work. When it rises, people do less checking and more exception handling. Both change cost and capacity.
When something goes wrong. The manager should know the fallback for each critical agent without looking it up, and know how to read the record to reconstruct any decision.
Withheld versus refused. Teams sometimes escalate a withheld action as a fault. The manager should explain that withheld actions are refused at the platform floor for every agent on purpose, and are not something to fix.
Controls and the evidence they leave
- Staggered calendar. Evidence: the calendar with dates and reasons for moves.
- Fallback plans. Evidence: approved plans per critical agent.
- Approver coverage check after each change. Evidence: dated check results.
- Team briefings logged. Evidence: notes and attendance per team.
- Practicum for managers who approve. Evidence: the practicum log template for operations managers.
What to measure
The number of weeks with two or more critical recertifications (aim for none). Briefings delivered within a week of each change. Shifts with fewer than two certified approvers for a depth 3 or 4 capability. And the age of the oldest held item each week, which shows strain earlier than any other figure.
Where does it go wrong?
Every agent was set up in the same month, so every certificate lapses together. Without deliberate staggering, one bad week can drop several agents at once.
Team leads skip the briefing because the change looks small. Small changes compound. The manager's weekly rhythm should list briefings outstanding.
Approver certificates lapse unnoticed. A queue with a recertified agent and no current approver is stuck. The manager answers for the operation as a whole, as are operations managers accountable for an AI agent's mistakes explains. Risk and property teams face related cycles; compare keeping risk managers ready and keeping property managers ready.
Where the Academy fits
Operations managers are a core audience for Pryme Intelligence Academy's Operator track, whose final module is running the rhythm. For teams, an admin view shows each person's progress, results, certificates and expiry dates, which feeds the 90-day view directly. Teams of up to 25 people can start with a free readiness assessment showing scores by concept and the gaps. More in the guide to AI literacy and certification for people who work with agents.
Questions people ask
How far apart should recertifications be staggered?
At least a week between critical agents, and never two agents that share a queue in the same week.
Should the operations manager attend every recertification?
No. They should see every result and every change note, and attend when a critical agent changes depth.
What if the fallback plan cannot be funded?
Then the agent's depth is carrying risk the organisation cannot cover. Raise it with risk before the next cycle.