Vocal
Sign in

Agentic Voice Agents based Mentorship and Counselling

Voice AI agents that work with campuses, not for them.

Built for counseling and mentorship desks. Vocal places the check-in calls and writes up what students said. Your staff keep the relationship and the next step.

Sign in

Calls, chat fallback, and a counselor queue in one app.

Counselors, mentors, and admins share one desk. Students never create accounts—they get a phone call or a one-time email link if the call is missed.

Example cohort

CSE · Year 2 wellness check-in

Outbound call

Call

A worker places paced outbound calls through Vapi. The assistant runs a short check-in with live language detection. Crisis phrases trigger a fixed escalation path, not a model judgment.

Concurrency limits

Voice session recorded for review

Missed-call fallback

Link

If the student does not answer after retries, a unique one-time chat link is emailed via Resend. The same check-in runs in the browser—speak or type. Each link is tied to one student.

Retries, then email

Unreachable only after call and link fail

Counselor queue

Queue

After each session, the AI writes a short report: who may need a human, why, and a confidence range. Flag reasons stay separate. Full transcripts are available on demand, not on the card.

Ranked by urgency

Counselor assigns, follows up, or dismisses

End result

A ranked list of students who may need a human, with flag reasons and confidence ranges attached. Transcripts are one click away.

What the AI writes after a check-in

The model reads the transcript and produces a short report for the queue. It flags patterns that may warrant follow-up. Crisis language is handled by a fixed branch, tested in eval fixtures—not left to the model.

Sample check-inReport output

Assistant asks

How has your week been? Anything making it harder to keep up with classes or feeling okay on campus?

Student says

Midterms were rough. I missed a few classes. I'd like to talk to Dr. Kavita if that's possible.

Queue entry

Attendance dropped after midterms. Student requested a counselor by name. Rank 01. Confidence 0.4–0.7. No clinical label on the card—just: a human should review.

Flag reason

Needs wellness follow-up

→

Asked for counselor by name · confidence 0.4–0.7

From cohort selection to counselor review

Staff enqueue weekly or event-based calls. BullMQ workers dial through Vapi with concurrency limits. Unanswered calls get retries, then a unique chat link. Crisis resources are always provided on-call and by email if a session drops.

  1. 01

    Select cohort

    Pick students to reach

  2. 02

    Enqueue

    Workers pace outbound calls

  3. 03

    Voice check-in

    Vapi assistant on the phone

  4. 04

    Retry

    Missed calls retried

  5. 05

    Chat link

    Unique link emailed if still unanswered

  6. 06

    AI report

    Patterns flagged, not diagnoses

  7. 07

    Human review

    Counselor takes next step

On each queue card

Rank, flag reasons, confidence range. Not the full transcript.

  • Ranked list of who needs attention
  • Confidence shown as a range
  • Separate flag reasons per student
  • Transcript available on click

Transcripts

Full text is behind a deliberate click.

Queue cards show rank, flag reasons, and a confidence range. Opening a transcript is intentional so the desk stays a triage list, not a scroll of student speech.

Staff sections

Reports, scheduling, events, and read-only query chat.

  • Reports

    Ranked attention list after each session

  • Schedule

    Enqueue recurring outbound calls

  • Events

    One-off cohort outreach

  • Query chat

    Ask questions about data—read-only, not a diagnosis tool

The AI does not assess, diagnose, or decide outcomes. It summarizes sessions and flags patterns. Crisis resources are provided on every call and emailed if a session ends abruptly. Students do not log in to this site.

FAQ

Common questions

Do students log in to this site?

No. Students receive a phone call or a one-time chat link. This web app is only for counselors, mentors, and campus staff.

Does the AI diagnose or assess students?

No. It summarizes what was said and flags patterns that may warrant a human look. A counselor, mentor, or admin decides follow-up.

What happens if a student mentions crisis language?

A hard-coded branch runs immediately—resources are spoken on the call and emailed if the session ends abruptly.

What if a student does not answer the phone?

The worker retries at a set pace. If still unanswered, a unique chat link is emailed. The student still appears in the queue once they complete the check-in or remain unreachable.