Cost optimization — CCA-F Exam Prep

PencilPrepPencilPrep
L2.34|Cost optimization
1/12
Person
A startup founder looking at two charts on her laptop. Left chart: 'AI Quality Score: 4.7/5 (stable).' Right chart: 'Monthly AI Cost: $22K → $4.8K.' She's smiling. Behind her, a whiteboard showing the optimization strategy: 'Route by complexity. Cache static prompts. Batch nightly reports.' Office, afternoon light.

Priya's AI bill was $22,000 a month. She cut it to $4,800 without losing a single point of quality.

Her customer service agent used Claude Opus for everything. 'What are your business hours?' -- Opus. 'Process this refund' -- Opus. 'Summarize this complaint' -- Opus. Every request, the most powerful and most expensive model.

She analyzed the requests. 60% were simple FAQ-type questions. 30% were moderate complexity. Only 10% needed Opus-level reasoning. She routed accordingly: Haiku for simple, Sonnet for moderate, Opus for complex. Added prompt caching. Batched the nightly reports.

Same product. Same quality. 78% less cost. She just stopped using a sledgehammer to hang picture frames.