Built for desks at

  • Counseling
  • Mentorship
  • CSE
  • ECE
  • AIML
  • Campus admin

VCampus counseling · Mentorship · AI notices, humans decide

Turn a quiet call into a human queue.

Vocal learns nothing clinical. It paces outbound voice check-ins, opens a missed-call chat, then ranks who needs a counselor. The AI never diagnoses.

Who should we reach?

Optional. Anything you write here is not stored — it just takes you to the desk.

Voice. Queue. Handoff. One desk, not three systems.

Counselors, mentors, and admins share the same rule: the model notices. A human assigns, follows up, or dismisses. Students never get an account.

Open cohort

CSE · Year 2 wellness check-in

Voice agent

Call

A short outbound check-in. Live language detect. No student app. Crisis language is a hard-coded branch, not a model whim.

Paced concurrency

A conversation, not a portal

Missed-call door

Link

If the phone goes unanswered, a unique chat link is emailed. Same check-in, typed. The link is not a shared inbox.

Retries, then email

Unreachable only after both fail

Staff desk

Queue

Counselors see who needs a human next. Confidence is a range. Flag reasons stay distinct. Transcripts stay one click away.

Ranked attention

A human takes the next step

One continuous system

People who need a human, with the evidence attached — not on the card.

A paced call to notice. A human to decide.

Vocal does not ask what a wellness form cannot hold. It asks, listens, and stops the moment crisis language appears. Eval fixtures treat that as pass or fail, not a vibe.

Paced check-inOne decision, made visible

Vocal

What would make this student need a human this week, not simply a missed class?

You

CSE year 2 after midterms. Quiet week. Two unanswered check-ins.

Notice · not diagnosis

Attendance dropped. The student asked for a counselor by name. Rank 01. Confidence 0.4–0.7. The card does not say “depression.” It says: a human should look.

Queue updated

Needs wellness follow-up

Asked for a counselor by name · range 0.4–0.7

One cohort in. The full check-in loop runs.

Enqueue weekly or event calls. Workers place them with concurrency limits. Unanswered attempts get a unique chat link. Crisis resources stay unconditional.

  1. 01

    Cohort

    Pick who to reach

  2. 02

    Pace

    Enqueue, don’t stampede

  3. 03

    Call

    Voice check-in

  4. 04

    Retry

    Missed? Try again

  5. 05

    Link

    Unique chat fallback

  6. 06

    Notice

    Patterns, not diagnoses

  7. 07

    Handoff

    A human decides

What reaches your desk

People who need a human.

  • Who needs attention
  • Confidence as a range
  • Distinct flag reasons
  • Transcript on request

Evidence

Transcripts stay off the card.

Rank, pattern, and a confidence range sit on the report. Raw text is a deliberate click — so the queue does not become a feed of other people’s words.

Operating

Six rooms. Same rule in each.

  • Reports

    Attention list after every session

  • Schedule

    Enqueue weekly calls

  • Events

    Cohort outreach

  • Query chat

    Read-only. Not a diagnosis.

The AI never assesses, diagnoses, or decides. It notices. A human takes the next step.

Crisis resources are unconditional — spoken on-call, emailed if a session drops. Students never log in here.

FAQ

Short answers.

Do students log in?
No. Students get a call or a one-time chat link. This site is the staff desk.
Does the AI assess or diagnose?
Never. It notices patterns and hands off. A counselor, mentor, or admin takes the next step.
What happens in a crisis?
Crisis language is a hard-coded branch. Resources are spoken and emailed if a session drops.
What if nobody answers?
Retries are paced. Then a unique chat link is emailed. The queue still ranks who needs a human.