Why your agent shouldn't quote prices

Language models are non-deterministic. Give the same pricing question to the same model twice and you can get two different answers. In a sales context, that inconsistency isn't just embarrassing — it's a legal and operational liability. A customer who was quoted one price by your agent and billed another has a legitimate grievance, and "the AI made a mistake" is not a defense that builds trust.

The fix is a hard boundary: the agent collects context, qualifies the lead, and hands off to a human or a deterministic pricing engine for the actual number. The LLM should summarize, not decide. Any action that has financial, legal, or irreversible consequences should be gated behind a non-LLM system.