GenAI Consumption & Cost Control

What you will take away

• Two cost surfaces, one governance gap: why consumption-priced API/token spend and per-seat AI licensing both need FinOps and SAM discipline at once.
• Unit economics that matter: how to reframe GenAI cost around cost per 1M tokens, per inference, per active user and per resolved outcome.
• The high-leverage levers: model right-sizing and routing, prompt caching, context discipline, commitment/PTU pricing and seat right-sizing.
• Licensing and compliance traps: data-use rights, IP indemnity, agent/message metering and per-user vs consumption double-pay.
• A worked benchmark: where 25–40% is typically recoverable on a representative ~$12M annual GenAI estate (benchmark scenario, not a quote).