Playbook
Our AI projects are blowing through budgets and we don't know which model is the cost
Several AI workloads on Azure OpenAI and Foundry are running over budget. The CFO can see the total but can't pin the cost on a specific model, a kind of prompt, or a team. Without finer-grained visibility, there's nothing to optimise.
Trigger — AI budget overrun; CFO demand for cost attribution.
Good outcome — Every AI workload's cost is split out by model, by team, and by prompt complexity. The conversation each month is "this model is expensive — is it the right one for the job?", not "something on AI is costing us money".