# Smart Escalation to Human Support | Adoption Layer

> AI resolves routine navigation and hands off complex requests to Chatwoot, Crisp or Intercom with the question, attempted steps and roadblock attached.

## AI guides the routine. Your team gets the exceptions, with the full story.

Adoption Layer walks users through your interface step by step. When a request needs a human decision (access rights, billing, policy exceptions), it hands the conversation to your support chat with the initial request, the completed steps and the blocking point already written.

## Tier-1 support is drowning in questions the interface could have answered.

Most tickets are not incidents. They are people who could not find a button, a setting or the next step. Each one costs an agent ten minutes of 'what were you trying to do, can you send a screenshot?'. Chatbots deflect by frustrating. Ticket forms arrive empty. Your specialists rebuild the context by hand before they can solve anything.

- **Navigation questions clog the queue**: 'Where do I export?', 'How do I add a user?': questions that need a pointer, not a specialist, still land as tickets.
- **Bots that dead-end**: Deflection bots stall on anything they cannot answer and leave the user to start over from zero in another channel.
- **Empty tickets, second round of questions**: Agents open a ticket with a one-line subject, no page, no history. Resolution starts with an interrogation.

## Every handoff arrives with the full story

The assistant writes the ticket the user never would. Three fields, filled from the live session, sent to the support tool you already use.

- **Initial request**: The question exactly as the user typed it, plus the page they were on. No more guessing intent from a subject line.
- **Completed steps**: Every guidance step the user followed before the roadblock, so agents never repeat what has already been tried.
- **Blocking point and reason**: Why the assistant stopped: out of scope, missing permission, rejected steps or a stalled session. The agent knows exactly where to start.

## Frequently asked questions

### Does the assistant replace my support team?

No. It handles navigation and 'how do I' questions inside the interface. Anything that needs a human decision, or anything it cannot resolve, is handed to your team with context. The goal is fewer, better-briefed tickets, not zero humans.

### Which support tools are supported?

Chatwoot, Crisp and Intercom today, with the chat opened and pre-filled at handoff. If no supported chat is present on the page, the handoff is sent to a fallback email you configure.

### What exactly does the agent receive?

A message built from your handoff template: the initial request, the completed guidance steps and the blocking point, plus the escalation reason and the session identifier so you can open the full session in the Adoption Layer dashboard.

### How do I define what the AI must not handle?

In the escalation settings you describe out-of-scope criteria in plain language (billing, refunds, security, legal...). The assistant applies them and offers a handoff on those requests instead of guessing.

### Can the user refuse the handoff?

Yes. The handoff is always an offer. The user can accept or dismiss it. Declined offers are tracked so you can see which requests users prefer to solve on their own.

### Is personal data sent to the support tool?

Only what the user typed and the structural context of the session. Sensitive values are masked in the browser before anything leaves the page, and hosting is EU-only. The conversation then lives in your support tool under its own privacy policy.