Genuine ambiguity
When a lead doesn't clearly qualify or disqualify against defined criteria, the honest move is to flag it for a person rather than force a decision the agent isn't equipped to make.
Explicit requests for a human
If a visitor asks for a person directly, that request is honored immediately — not deflected with another round of automated questions.
High-value or high-complexity inquiries
Large deal sizes or unusually complex requirements are routed to a person once qualification confirms genuine fit — see Websites for High-Ticket Services for why this matters more as deal size grows.
Anything outside the defined knowledge boundary
If a question falls outside what the agent was built to know, it says so and hands off rather than guessing — see How to Prepare a Knowledge Base for an AI Agent for how that boundary gets set.
How this actually gets designed and tested
Handoff triggers are written down explicitly during setup — not left to the agent's discretion — and tested deliberately before launch by forcing each trigger condition on purpose. See How to Test an AI Agent Before Launch for exactly how that testing works, and How to Handle AI Agent Failures and Retries for the related question of what happens when the agent misjudges before a handoff trigger fires.
Questions
Does the AI ever pretend to be human?+
No — it's presented honestly as an AI agent, not disguised as a person.
What happens during handoff if no one is available immediately?+
The lead is logged and flagged, with an honest message to the visitor about response time — not a silent gap.
Can handoff rules be customized per business?+
Yes — the specific triggers are configured against your business's actual qualification criteria, not a fixed generic list.
Are handoff triggers tested before launch, or just assumed to work?+
Tested deliberately — each trigger condition is forced on purpose during pre-launch testing, not assumed to work from design alone.