Overview
How might we design ai-powered customer support so people can trust and act on AI output?
When to use
- Perfect for e-commerce platforms, SaaS applications, and customer service tools where proactive AI support improves satisfaction and reduces support costs.
When to skip
- Crisis incidents where only human agents should respond.
- Proactive chat that blocks checkout on every error.
- Issues requiring identity verification you cannot perform.
Rules
Bot loops with no talk to human within two failures.
Proactive popups on every page view.
Wrong order context in support suggestions.
Closing tickets while user still sees the problem.
Evidence
| Product | Implementation |
|---|---|
| Amazon | Where is my stuff and issue-specific help hubs. |
| Zendesk | Answer Bot and intent routing before agent queue. |
| Intercom | Proactive messages from product and billing events. |
| Drift | Playbooks that qualify and route support chats. |