Halving the LLM bill behind WhatsApp chatbots
Indigo Hive, 2026 – now
Monthly LLM API cost
- Before
- 100%
- After
- 50%
−50%
LLM API costs
- The problem
- AI-driven WhatsApp chatbots run on paid LLM APIs, so every prompt and every token shows up on the bill.
- My role
- Software engineer on the AI workflows and their cloud infrastructure.
- What I built
- Rewrote prompts and cut token usage per request.
- Made the AI request flow more efficient.
- LLM-backed workflows and integrations running on AWS.
- Outcome
- −50% LLM API costs.