Agentic UX
After the agent acts

Escalation Pathway

When the agent is unsure or the stakes are high, it pauses and brings in a human with context.

When to use

Agents that guess on edge cases create risk and frustration.

Example scenario: Support bot hits billing edge case; loops in human with ticket ID in-chat.

Anatomy

UI pieces that make this pattern recognizable.

  • Clear pause message and reason code
  • Warm handoff with transcript and collected fields
  • Human agent visible in same thread or inbox

Guidance

Do

  • Distinguish 'offer human' vs 'immediate handoff'
  • Never blame the user for hitting an edge case

Avoid

  • Do not loop infinitely on 'let me try again'
  • Do not drop conversation context on escalation

Limitations

When this pattern adds friction or fails to help.

  • Over-escalation burns human capacity and signals weak automation
  • Under-escalation on high-stakes domains creates liability
  • Handoffs that drop structured fields force humans to re-ask questions

Build notes

Implementation hints for engineers shipping the pattern.

  • Configurable escalation rules + natural-language guidance
  • Route to queue with priority based on intent and sentiment

Examples

Annotated screenshots from production products, with designer critique.

Production screenshots for this pattern are not in the gallery yet. Use the Guide to audit your product, or check the Crosswalk for related catalogues.