THE PLATFORM

Run every campaign from one pane of glass.

AVA does the calling. The platform is where your team watches it happen, proves every word it said, and steps in the moment a lead turns hot. One workspace — from the lead landing to the counsellor closing — with nothing spoken that a human hasn't approved.

app.eduvy.in · one tenant, fully isolated · India-hosted

Grounded, zero guesses DLT-compliant DPDPA-ready

The Eduvy command center — live calling operations across every campaign — Command Center — live dialer ops: calls in flight, the needs-you list, live ₹ cost-per-CLI

01 — Ground it first

Nothing is spoken until a human says yes.

Before AVA dials a single lead, it has to earn the right to speak. Upload a prospectus or point at a URL; extraction turns it into draft facts — each fee, date, eligibility rule and scholarship with its source location and a confidence score. Your team reviews them in a queue, edits the exact wording, and approves. Only approved, in-date facts can ever leave AVA's mouth — and it speaks numbers and dates verbatim, never rounded, never guessed.

DLT-compliance gate KB-readiness gate

Draft → Review → Approved
bulk-approve high-confidence drafts
warns before a fact lapses · Expiring ≤30d
both gates must pass before launch · DLT ✓ · KB-ready ✓

02 — Put it to work

Outbound calling, on your terms.

Build a campaign in a few clicks: name it, pick the target leads by filter or hand-pick them, set the calling window and days, retries, and a budget cap — then launch. AVA dials matched leads automatically, but only inside your calling-hour windows, only after DNC scrubbing, and never past a budget you set. A/B different openings to learn what gets consent fastest. Pause, resume, add leads, or place a single test call before it runs at scale.

concurrency cap [—]
calling window 09:00–21:00 IST (set per campaign)
budget cap ₹[—]
retries on no-answer [—]

Launch stays locked until both gates from Station 1 read green.

03 — The war room

Watch every line, the second it connects.

This is live calling operations across all your campaigns at once. Campaigns running now and their progress, calls in flight this minute, the "needs you" list of callbacks and retries waiting on a human — and, ticking in real rupees, the cost per caller-ID as it dials. Per-CLI number health sits right beside it, so a number starting to get spam-flagged is visible before it costs you answer-rate. Calling-hour windows, DNC scrubbing and budget caps are enforced here, not hoped for.

live cost-per-CLI ₹[—]/CLI ticks live · per-tenant figure, never invented

calls in progress [—]
calls/min [—]
per-CLI answer-rate [—]%

Command Center — live dialer ops, calls in flight, per-CLI number health — Command Center — the war room: campaigns now, calls in flight, the needs-you list, live ₹ cost-per-CLI

04 — Know who's hot

Every lead, scored — and the cold ones, caught.

Each lead carries a live score from 0 to 100, banded Hot / Warm / Mid / Cold, filterable by source, program, stage and owner — with saved views your team reuses every morning. The Dashboard is the day's first read, with an AI daily briefing that tells you what changed overnight. The one that earns its keep is the going-cold radar — high-intent leads with stale contact, ranked by decay risk, with a one-click "re-engage all." A high score going quiet is the most expensive thing in admissions; the radar makes sure it never goes unseen.

score 0–100 · bands Hot · Warm · Mid · Cold
hot = score ≥ 85
going-cold radar ranks by decay risk (score × staleness)

05 — The whole student

One record. Every word. Why the score.

Open any lead and the whole relationship is on one page: the score gauge with the "why this score" reasons spelled out, the qualification profile AVA captured on the call, the application-stage funnel, the full interaction timeline, call attempts with transcripts, the WhatsApp thread inline, and the tasks and callbacks attached. This is conversation intelligence, not a contact card — your counsellor walks into the call already knowing the student. And "Call" dials them with AVA right now.

0 score · why this score: Fit · Intent · Freshness

captured fields
program · intake · marks
location · timeline · budget signal

06 — The anti-hallucination trail

AVA can prove every number it ever said.

This is the screen that lets you sleep. For any past call, open the grounding audit: every answer AVA gave, and the exact approved facts that grounded it — the source, the approved stamp, the in-date check. Grounded answers carry their Cited facts; deferrals are marked with their reason. A stale fee can't be quoted, because retrieval hard-filters on approved-and-in-date at the database, not on a guess. And the per-call COGS in ₹ ticks up right beside the audit, so accuracy and economics live on one screen.

“B.Tech CSE ki fees kitni hai is saal?”

Grounded · fee · B.Tech CSE

Cited fact #a1b2c3d4 approved · in-date · spoken verbatim

“Management quota mein seat mil jayegi kya, aur kitne mein?”

Deferred · no approved, in-date fact

forked aside, dimmed to amber — a counsellor will confirm, rather than guess.

per call: [—] grounded · [—] deferred
Cited facts: #a1b2c3d4 …
per-call COGS ₹[—]

07 — It gets sharper every week

Every question AVA couldn't answer becomes your next move.

When AVA defers, it doesn't just protect you — it learns. Every deferred question is logged as a knowledge gap and ranked by how often it's asked, turning "what we couldn't answer yet" into a prioritised worklist. Answer the top one in the Knowledge Base and AVA can ground it on the very next call. The same view shows conversion by source, program demand, and caller-number answer-rate health. This is the self-improving knowledge-base loop: the gaps your prospects actually have, surfaced and closed, week over week.

gaps ranked by asked [—]×
feeds: conversion-by-source · program demand · number health

deferrals flow back up the filament into the KB — the loop closes.

08 — And everything around it

The rest of the room.

Four more surfaces complete the operation — design the voice, manage the messages, see the true cost, and route the humans.

Voice Studio

Design how AVA sounds and speaks: persona, voice, the languages it offers, and every spoken line — greetings and hand-offs — with auto-translation into your other supported languages and previewable samples. The one thing you can't edit is the grounding rules: figures always come from approved facts.

9 Indian languages + Hinglish · greetings · hand-offs · grounding locked

WhatsApp templates

Per-college templates you approve yourself, opt-in management, and messaging gated by opt-in and the 24-hour window. Inbound WhatsApp is answered by the same grounded AVA, with the same defer rules — and the thread shares context with the call.

opt-in gated · 24h service window · omnichannel memory

Billing / COGS

Transparent per-call economics, metered from each call's real data — broken down across Telephony, STT, TTS and an estimated LLM line — with the daily cost trend, the rate card, and a per-call cost lookup. No mystery markup; you see where the money goes.

Telephony · STT · TTS · LLM~ · COGS (30d) ₹[—] · avg/call ₹[—]

Counsellor queue

Hand-offs and callbacks, most urgent first — every escalation AVA created when it deferred, a caller asked for a human, or a callback was scheduled. Counsellors claim, work, and close. The sidebar badge keeps the open count honest.

escalations first · [—] open · claim → work → done

→ In their pocket

And in your counsellors' pocket —

The command center runs the operation. When AVA hands a hot lead off, the close happens on a phone.

→ The Counsellor App