> For the complete documentation index, see [llms.txt](https://oodlz-ai-studio.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://oodlz-ai-studio.gitbook.io/docs/inbound-agent/the-knowledge/human-escalation.md).

# Human Escalation

Full conversation context handed off to support in seconds — branded email to both parties, validated payload, white-label by default.

> **Real person requested, full context delivered — no cold handoff, no repeat explanations.**

Most AI systems either never escalate to a human, or do it badly — dropping the conversation context and making the user explain everything again from scratch. The Oodlz human escalation workflow handles the handoff in seconds. Branded confirmation email to both the support team and the user. Full conversation context attached. Reference ID issued so the user knows the escalation is real.

## What You Get

* One-call escalation — the agent collects email and issue description, then fires the [handoff](/docs/inbound-agent/the-knowledge/human-escalation/the-handoff.md) in a single tool call.
* Branded confirmation email to support and the user simultaneously — both parties know the request is in flight.
* Full conversation context handed off — session reference, original query, issue description, all in the email body.
* White-label by default — every brand element pulled from the brand profile at runtime; no hardcoded Oodlz branding anywhere.
* Reference ID returned to the user so they have something to quote in any follow-up.

## How It Works

When the agent classifies a message as needing human escalation — `negative_complaint`, explicit "speak to someone" request, or persistent frustration — it collects the user's email and a short issue description in conversation. Once both fields are validated, the agent calls the `human_escalate` tool exactly once.

The escalation pipeline reads the brand profile at runtime, builds a branded HTML email, and sends it via Gmail in a single call to both recipients: the brand's support email and the user's email. The body includes the user's email, the issue description, the session reference ID, and the original query that triggered the escalation. Support gets the full context they need to respond. The user gets immediate confirmation that the request was received.

The pipeline returns a structured success response to the agent. The agent then closes the loop with the user — confirming the escalation completed and citing the reference ID. The entire handoff happens in seconds with no human intervention until a support agent opens their inbox.

## Quality Baked In

* Email and issue description validated by the agent before the tool fires — support never receives a blank or half-formed escalation.
* Single Gmail call to both recipients — the support email and the user email arrive simultaneously, no race between them.
* Session reference travels in the email body — support can look up the full conversation history before replying.
* Every brand element pulled from the brand profile — no Oodlz strings hardcoded in the email template, the workflow ships ready for any brand.
* Structured success response back to the agent — the agent knows the escalation completed before confirming to the user.

## Inputs

* `human_escalate` tool call from the inbound agent.
* 4 structured fields: user email, issue description, session reference, original query.
* Brand profile fields — name, support email, logo URL, website URL — pulled at runtime.

## Outputs

* Branded HTML email delivered to the brand's support team and the user.
* Structured success response back to the inbound agent.
* Reference ID communicated to the user as confirmation.

## What This Replaces

* The "fill out this form, someone will get back to you" page that loses conversation context entirely.
* The support inbox flooded with escalations that have no thread, no history, no reference.
* The follow-up email asking the user to explain what happened from the start.
* The hardcoded brand strings that have to be edited every time the workflow ships to a new client.

One tool call, both parties notified, full context attached, no hardcoded branding — every escalation, every brand.

***

*This was built. No agent ships this out of the box.*