Can LizziAI escalate complex issues to a human?

Yes — escalation to a human is a first-class feature in LizziAI, with configurable rules per category, client, and severity. You define when LizziAI should hand off; LizziAI decides at runtime if a given conversation crosses that threshold and routes it to the right person with full context.

Built-in escalation triggers

• Sentiment shift — angry, frustrated, or threatening language detected
• Topic sensitivity — pricing, legal, complaints, refunds, churn signals
• Confidence threshold — LizziAI flags when it's <80% confident in its proposed action
• VIP client tagging — high-value clients always get human-first treatment
• Out-of-scope requests — anything outside the agent's defined responsibilities
• Repeat asks — if a client asks the same question twice, escalate
• Manual keyword triggers — phrases like "speak to a manager" or "cancel my account"

How escalation works

1. LizziAI flags the conversation in the unified inbox with an escalation badge and reason
2. The conversation is auto-routed to the assigned owner (or a fallback queue)
3. The owner gets a notification via email, SMS, or in-app chat — your choice per user
4. LizziAI provides a contextual brief: "Acme is asking about a refund for the Q1 retainer. They've been a client for 14 months. Last invoice paid on time. Suggested response below — review before sending."
5. The human can edit, approve, or take over completely

Configurable autonomy

For each category of work — appointment confirmations, billing questions, technical support, complaints — you set the autonomy level: full autonomy, draft-and-approve, or human-only. New tenants start in conservative mode (everything draft-and-approve) and graduate as you trust LizziAI more.

SLA-driven escalation

If a human owner doesn't respond within your defined SLA (e.g., 4 hours), LizziAI re-escalates to a backup owner or alerts a manager. No client gets stranded. Request beta access to configure escalation for your team.