Azure Cognitive Services (AI Services) Account
Does ZopNight manage Azure Cognitive Services (AI Services) Account?
Azure AI Services accounts meter per API call on pay-as-you-go SKUs, or through prepaid commitment tiers that bill the committed amount regardless of traffic. ZopNight inventories every account via Resource Graph with kind and SKU detail, expands OpenAI-kind accounts through the data plane, and attributes usage and commitment spend from Cost Management.
Rules that fire on Azure Cognitive Services (AI Services) Account
At a glance
| Field | Value |
|---|---|
| Scheduling notes | discovery and cost visibility only. |
Azure AI Services accounts provide APIs for vision, speech, language, and OpenAI models, billed per call or via commitment tiers. Committed capacity purchased for pilots that ended keeps billing regardless of traffic.
Two meters: per-call and prepaid commitment
An AI Services account bills in one of two modes. Pay-as-you-go SKUs meter each API transaction, whether an image analyzed, an utterance transcribed, or a document translated, so cost tracks usage and an idle account costs nothing. Commitment tiers invert the deal: a fixed monthly amount buys a discounted block of capacity, and that amount bills whether the workload consumes all of it, some of it, or none of it. The discount is real when traffic is steady; the same mechanism becomes a standing loss when traffic falls away and nobody revisits the tier.
What kind and SKU reveal at discovery time
ZopNight discovers every account through Azure Resource Graph, carrying its SKU and its kind, which names the service family the account exposes. Kind determines what happens next: accounts of the OpenAI kind get expanded by the OpenAI enricher into their deployments, jobs, and vector stores, while classic vision, speech, and language accounts are tracked at the account level. Cost Management billing attributes both per-call usage and commitment spend, so a committed account’s fixed fee and its actual consumption can be compared. Accounts are discovery-only. There is no stop verb on an API endpoint, so findings here feed review rather than schedules.
Commitments that outlive their pilot
The signature waste on this type: a speech-to-text or translation pilot buys a commitment tier for the discount, the pilot wraps, and the tier keeps renewing against near-zero traffic. A second pattern is fragmentation: many single-purpose accounts scattered across resource groups, each too small to notice, together forming a meaningful monthly floor.
Surveying accounts across a subscription
Azure portal → Azure AI services lists every account with kind, SKU, and region; sorting by pricing tier is the quickest way to spot committed accounts worth an honest utilization check.