Skip to main content

Mental health and safety escalations: How your AI One Stop agent responds

This article is intended for Student Affairs leadership and the staff who will receive escalation notifications. Here, we’ll explain how default behavior of your AI One Stop agent when a conversation involves mental health, distress, safety, or student wellbeing. This behavior may be customized to best suit your institutional needs and best practices.

A reminder, that ‘escalation’ means that an email will be generated to your preferred contact, notifying them of the conversation in question. This does not include a live call transfer. For more on escalation handling in CollegeVine, please see this article.

The core commitment

Your agent is designed to bias toward escalating whenever wellbeing is at stake. Missing an escalation is treated as a critical failure. Triggering an unnecessary one is treated as a few minutes of staff time. The calculus always favors getting a human involved.

What triggers an immediate escalation

The agent escalates the moment it detects any of the following signals. No delay, no clarifying questions, no asking the student to repeat themselves.

  • Crisis language or self-harm indicators

  • References to harming others or threats to campus safety

  • Distress signals, including hopelessness, despair, or being overwhelmed

  • Concern expressed about another student's wellbeing

  • Disclosures of food, housing, or basic-needs insecurity

  • Discrimination, harassment, or Title IX concerns

  • Mid-term withdrawal or leave requests paired with wellness signals

  • Explicit requests to speak with a counselor, advisor, or specific staff member

  • Legal questions, threats, or references to attorneys

The agent uses behavioral interpretation, not a keyword list. It evaluates the full conversation, including context and tone, not just specific words. This matters because students rarely use clinical language. A student who says "I just can't keep going like this" receives the same response as a student who uses explicit crisis language.

Two principles behind these triggers

  1. Over-escalation is acceptable. Under-escalation is not.

  2. Escalation is not failure. The agent treats it as normal routing, not as an inability to help. It will not apologize for escalating, frame it as giving up, or imply the student did something wrong by raising the concern.

What the student experiences in the moment

In a safety or wellbeing scenario, the agent does several things at once:

  • It continues the conversation rather than going silent. Students are not abandoned mid-crisis.

  • It directs the student to a real, immediate human resource (campus safety, a counseling line, or whatever number you configure for distress situations).

  • It emphasizes that the agent is not a crisis resource and the student should reach a person directly.

  • It never claims to be capable of providing mental health support, counseling, or safety intervention.

  • It does not tell the student an escalation was triggered behind the scenes. This is deliberate. It prevents two failure modes: students "trying to trigger" the agent for attention, and students worrying about whether their message actually reached someone.

What your team receives

Every escalation generates a notification to the email address you configure for that escalation type. The notification includes:

  1. The student's name and contact information, auto-pulled from their record (the agent does not ask the student to provide this)

  2. The channel the conversation happened on (web chat, email, SMS, or phone)

  3. A summary of the concern, including a topic label

    1. If the escalation as a result of a phone conversation a link to the full call transcript is provided

  4. What the agent already said to the student

  5. Why the situation requires human attention

  6. Urgency level. Safety and mental health escalations are always flagged as critical, meaning immediate attention required.

The agent uses the full record on file, so your team receives the student's actual contact details, not whatever they typed into a chat form. If your CRM is integrated, student type and program designation also come through.

What you control

These decisions are made during onboarding and live in your agent's configuration:

  • The destination email for safety and mental health escalations. This can be the same inbox as your general escalations or a separate inbox dedicated to crisis routing. CollegeVine refers to a dedicated route as crisis_support.

  • The exact phrasing the agent uses with students in distress. Some partners want the agent to share a specific number (Department of Public Safety, a 24/7 counseling line, an after-hours service). Some want the agent to name a specific office or resource. These customizations are applied to the agent's prompt.

  • Additional triggers beyond the defaults. If your campus has specific scenarios that should always escalate (mentions of a specific medication, a specific building, a specific event), those go in the workbook.

What is fixed, by design

A few elements are not configurable because they are part of the agent's core safety posture:

  • The agent will always escalate on the triggers listed above. These cannot be turned off.

  • The agent will always direct students to a human resource in a crisis. It will not attempt to provide counseling itself.

  • The agent will never tell a student that an escalation has been performed or imply that someone is on the way. It does not give false reassurance about backend actions.

Recommended setup

Based on what works well across other AI One Stop deployments:

  1. Set up a dedicated crisis_support route to a monitored inbox or distribution list. Routing crisis traffic to a general support inbox slows response. A dedicated, frequently monitored destination ensures the right people see the right escalations fast.

  2. Configure the agent to share a specific safety contact in distress responses. Whether that is your Department of Public Safety, a 24/7 counseling line, or another resource your team trusts, this gives students a concrete next step beyond "reach out to a counselor."

  3. Decide as a leadership team who monitors the safety escalation inbox after hours. The agent is available 24/7. Make sure the destination for safety escalations is too, even if monitoring shifts between offices.

  4. Walk through the actual response language together before launch. During your testing, look at exactly what the agent will say in a distress conversation. Adjust it for your campus voice. Make sure it sits right with the people responsible for student welfare.

What the agent is, and is not

The agent is a routing mechanism, an information source, and a 24/7 first point of contact. It is not a counselor, a clinician, or a substitute for the trained professionals on your campus.

Its job in a mental health or safety scenario is to get a human involved as fast as possible, give the student direct guidance on how to reach immediate support, and pass complete context to your team. The agent will never claim to know how a student is feeling, provide therapeutic advice, or attempt to assess clinical severity. Those responsibilities belong with the people on your team. The agent's job is to make sure those people know, fast, that they are needed.

Questions about the escalation framework, custom triggers, or routing setup go to your CollegeVine Implementation Manager or VP of Partner Success.

Did this answer your question?