AI Commitment Plan
Does ZopNight manage AI Commitment Plan?
An AI commitment plan is a prepaid throughput reservation under a Cognitive Services account, billed as capacity units times hours for the committed amount whether traffic arrives or not. ZopNight reads each plan's type, tier, and unit count, along with a hosting model such as ProvisionedManaged, and compares committed spend against actual usage.
Rules that fire on AI Commitment Plan
No active rule family targets AI Commitment Plan today. Rules that used to are retired, and retired rules publish no pages and fire no findings. Scheduling and permissions coverage are unaffected.
At a glance
| Field | Value |
|---|---|
| Scheduling notes | discovery and cost visibility only. |
Commitment plans lock in discounted AI Services capacity for a fixed monthly fee. A plan sized above actual usage silently pays for capacity nobody consumes.
Prepaid capacity units, metered by the hour
A commitment plan is the purchase record behind provisioned AI capacity: a prepaid reservation attached to a Cognitive Services account, billed as capacity units multiplied by hours for the full committed size. The meter reads the commitment, not the traffic. A plan running at a fraction of its capacity bills exactly like one running flat out. That is the intended bargain (a discount in exchange for predictability), and it is also why this type deserves its own inventory row: the plan is directly billable on its own, separate from any per-call usage the account records.
The plan fields that drive a rightsizing review
ZopNight’s AI capacity enricher lists commitment plans under every Cognitive Services account, a sub-resource that Resource Graph does not surface on its own, and records the fields that matter for cost: the plan type, the active period’s tier, the unit count of the current commitment, and the hosting model, such as ProvisionedManaged. Cost Management then sets committed spend beside actual usage, turning “we have a commitment” into “we use 30 percent of it,” which is the sentence a rightsizing decision actually needs. Plans are discovery-only; changing a commitment is a contractual action taken by a human, so the platform’s role ends at making the mismatch undeniable.
Where committed AI capacity goes stale
Two shapes recur. A plan sized from optimistic launch forecasts that real traffic never met, renewing period after period at the original count. And a workload migration, often to newer OpenAI deployments under a different account, that leaves the old plan committed to a service nothing calls anymore.
Checking a plan against its account
Azure portal → the Cognitive Services account → Commitment tier pricing shows the active plan, its tier, and its renewal setting; compare it with the account’s metrics before the next period rolls over.