Cost Optimisation¶

Concrete techniques and a cost estimation tool for AI systems. Use this document in the Development and Monitoring & Optimisation phases to keep costs manageable.

1. Cost Estimation (Calculator)¶

Complete the table below for a quick monthly estimate.

LLM API Costs¶

Monthly input costs  = (requests/day × 30 × input tokens) / 1,000,000 × price
Monthly output costs = (requests/day × 30 × output tokens) / 1,000,000 × price
Total API costs/month = input costs + output costs

Example: 500 requests/day → 500 × 30 × 800 / 1,000,000 × €2.50 = €30/month input + 500 × 30 × 300 / 1,000,000 × €10 = €45/month output = €75/month total

Total Monthly Cost Estimate¶

Scenarios:

2. Optimisation Techniques¶

Technique 1 — Prompt Optimisation¶

Expected saving: 20–40% on input tokens

Unnecessary tokens in system prompts and user instructions increase costs without quality gains.

Technique 2 — Response Caching¶

Expected saving: 30–60% for repetitive queries

Identifiable, repeated questions (FAQ, standard reports) are cached rather than re-sent to the API.

Technique 3 — Model Tiering¶

Expected saving: 40–60% for mixed workloads

Not every question requires the heaviest (most expensive) model. Route based on complexity.

Example routing logic (Python):

def select_model(query: str, complexity_score: float) -> str:
    if complexity_score < 0.3:
        return "claude-haiku-4-5-20251001"   # Light — fast & cheap
    elif complexity_score < 0.7:
        return "claude-sonnet-4-6"           # Medium — balanced
    else:
        return "claude-opus-4-6"             # Heavy — complex reasoning

Technique 4 — Chunking & RAG Optimisation¶

Expected saving: 20–40% on context length for document processing

Technique 5 — Batch Processing¶

Expected saving: 30–50% for non-real-time workloads

• Use Batch API endpoints (Anthropic, OpenAI offer 50% discounts)
• Schedule heavy processing outside peak hours
• Combine multiple documents in one API request where possible

3. Monitoring & Cost Management¶

KPIs for Cost Management¶

Budget Alert Configuration¶

Always configure budget alerts at:

• 70% of monthly budget → warning notification
• 90% of monthly budget → escalation to AI PM + CAIO
• 100% of monthly budget → automatic rate limiting or stop

Cost Allocation¶

Allocate costs per system, team or use case via tags/labels in your cloud environment. This enables ROI calculation per project (see Benefits Realisation).

4. Cost Optimisation per Phase¶

Related Modules¶

• Cloud vs. On-Premise
• MLOps Standards
• Benefits Realisation
• Business Case Template
• Agentic AI Engineering — Cost Management
• Engineering Patterns