—
total per month—per resolved ticket
—saved vs all-human
—AI share of cost
Deflection, not token price, decides the ROI
AI handles the easy tickets for cents, but every escalation still lands on a human at their loaded cost. Push the deflection rate and watch the blended cost per ticket fall; drop it and automation can cost more than the team it was meant to save. Model the underlying bot on the chatbot cost calculator, size the context with the RAG cost calculator, and pick the cheapest capable model with the cheapest LLM API tool.