Admissions AI for an Education Academy

The challenge

The academy — 2,000 students across school-support, competitive-exam and professional programmes — had a marketing engine that worked and an admissions funnel that leaked. Enquiries arrived through web forms, WhatsApp, phone and walk-ins at a rate of several hundred a month around intake season. The admissions team of four responded in order, which meant the average enquiry waited 26 hours for a first reply; evening and weekend enquiries — 62% of the total, as it turned out — waited longest. Follow-up lived in notebooks and memory. Management knew conversion was poor but couldn’t even measure it precisely, because the funnel had no single source of record.

In a city where families typically shortlist three or four institutions in an evening, the first useful response often wins. The academy was systematically losing that race.

The solution

We deployed a conversational AI admissions assistant across the academy’s website and WhatsApp — WhatsApp being where Kerala families actually communicate. The assistant answers the factual layer instantly and around the clock: programmes, eligibility, fees, batch timings, faculty, transport. It asks qualifying questions naturally in the flow of conversation, scores each enquiry for programme fit and intent, writes everything to the CRM, and books counselling appointments straight into counsellors’ calendars for high-intent prospects.

Behind the conversation sits an agentic follow-up layer: every enquiry gets a next action and an owner, nudges go out on schedule and adapt to responses, and no lead is ever silently dropped. Counsellors start their day with a ranked queue and full conversation context instead of a pile of sticky notes.

The architecture

The conversational layer runs on an LLM grounded in a curated knowledge base the academy’s team edits directly — fees and batch schedules change often, and content control had to sit with staff, not with us. Guardrails keep the assistant inside admissions scope: it never improvises fee discounts, never gives academic-outcome guarantees, and hands over to a human on request or on low confidence. The scoring model is deliberately simple and auditable (programme fit, urgency signals, engagement depth), and the orchestration layer treats the CRM as the single source of truth. Conversation logs are retained with consent notices appropriate to Indian data-protection law.

The results

Within one intake cycle, enquiry-to-enrolment conversion tripled. First response time fell from 26 hours to under a minute, and — the number that surprised the academy most — 62% of enquiries were arriving outside office hours, demand the old process had structurally ignored. Counsellor time shifted from repetitive information-giving to actual counselling, and management gained a live funnel dashboard where before there had been a monthly guess. The system paid for itself within the first intake on recovered enrolments alone.

What this means for your institution

Admissions is a sales funnel with a conscience, and it responds to the same physics: speed, persistence and context win. This deployment also seeded our own AI Academy in Kochi — we teach with the systems we build. If your institution measures enquiries in hundreds, book a strategy call; bring one month’s enquiry log and we’ll estimate your recoverable enrolments on the call.