For each client billed per seat or per device, establish what the count actually is this
month — from RMM device inventory and onboarding/offboarding tickets — and compare it to the
count being billed, producing a true-up evidence pack: who is under-billed, who is
over-billed, and the tickets/devices that prove it. (For per-license SKUs like M365/AV/backup,
use License Billing Reconciliation instead.)
1. Confirm scope — all per-seat/per-device clients or one client — the billing month, and the
source of the BILLED counts. Billed quantities live in the billing system, not Thread: ask
the requester to paste or upload the current billed count per agreement. No billed counts =
no true-up; offer the actual-count half of the pack alone, clearly labeled.
2. Per client, establish the actual count from the strongest available sources: pull the
device inventory from the RMM this tenant has (NinjaOne or ConnectWise RMM), filtered to
billable device classes per the agreement (exclude spares/decommissioned; state the filter
used).
3. Corroborate with movement tickets: read the onboarding/new-user and offboarding/termination
tickets for the month. Movement should explain the delta between last month's and this
month's counts; unexplained deltas get flagged rather than smoothed over.
4. Per agreement, compare actual vs billed: delta, direction (under-billed = revenue leak;
over-billed = client-trust/credibility risk — flag both with equal seriousness), and the
monetary impact if the per-unit rate is provided (labeled with the rate used).
5. Build the evidence pack per discrepancy: the device list or count with source and as-of
date, the onboard/offboard ticket references, and a one-line narrative ("3 users onboarded
in tickets <refs>, never added to billing"). Evidence must survive the client asking "prove
it."
6. Output: a true-up table (client / billed / actual / delta / direction / monthly impact),
the per-discrepancy evidence packs, a recommended adjustments list for the billing admin,
and a methodology note — sources per client, device filters, as-of dates, searches run,
caps hit.
Guardrails: this skill produces the evidence pack; it never changes billing, quantities, or
invoices — never convert a recommended adjustment into a completed one. Over-billing is
reported as prominently as under-billing (a true-up that only finds revenue is not credible).
Device counts are as-of-date snapshots with stated filters; a stale RMM agent is not a
billable seat — flag devices unseen for 30+ days separately. Never present capped device or
ticket searches as complete counts; disclose caps and re-split until counts are exact, since
billing math cannot run on floors. If no RMM integration is enabled, say so and fall back to
movement tickets only, labeled a weaker count basis. Monetary impacts use only rates the
requester supplied — never assumed per-seat pricing.
