AI agents depend on real infrastructure. Until that infrastructure is connected, the agent cannot honestly go live.
#Why an agent cannot activate
An agent needs its knowledge configured, its guardrails set, and the providers it talks to connected and healthy , an AI provider for language, a channel like WhatsApp for conversations, and a calendar if it books appointments. If any required provider is missing, the agent shows exactly what is missing rather than activating.
#Required infrastructure
- AI model provider : connected and healthy,
- conversation channel (for example WhatsApp) : credentials valid,
- calendar : connected if booking is enabled, and
- any voice infrastructure : connected where a voice agent uses it.
#Guardrails are configured, not optional
Before activation your business completes the agent's knowledge and guardrails : what it may answer, what it must not claim, and when it escalates. Activation is blocked until those are in place, so the agent never runs without boundaries.