Admissions Platform: AI Admissions Assistant
Overview
A study-abroad admissions consultancy was answering the same questions about courses, countries and eligibility by hand, one enquiry at a time. We built an LLM-powered admissions assistant into their new website that answers from an approved knowledge base, works out what each student is actually asking, and captures the details a counsellor needs. Conversations that show real intent are handed to a counsellor with the context already attached, so the first human reply starts where the chat left off.
Assistant availability
Enrolments with a smaller team
Approved knowledge base behind every answer
Qualification fields captured per enquiry
The challenge
we had to solve
Prospective students arrive with open-ended questions (which country suits their budget, whether their grades meet entry requirements, when intakes open) and expect an answer immediately, often outside office hours. The consultancy had no way to respond instantly, and every enquiry reached a counsellor as an unstructured message that had to be re-asked from scratch. A generic chatbot was not an option: in admissions, a confidently wrong answer about eligibility or deadlines damages trust and wastes counsellor time.
How we
approached it
We grounded the assistant in a curated knowledge base of the consultancy's own approved content, using retrieval so answers are drawn from vetted material rather than the model's general knowledge. An intent-classification layer separates browsing questions from enrolment intent, and the assistant gathers structured fields (destination, course level, intake, academic background) as part of a natural conversation, with consent captured before any contact details are stored. When an enquiry qualifies, or a question falls outside the approved content, it escalates to a counsellor with the full transcript and a structured summary.
Grounded Answers
Retrieval over an approved knowledge base keeps answers on courses, countries and eligibility tied to content the consultancy has signed off.
Intent Classification
Each conversation is classified by what the student needs (general information, eligibility check or ready to apply), so the assistant responds at the right depth.
Structured Capture
Destination, course level, intake and academic background are collected conversationally and stored as clean, structured enquiry data with consent recorded.
Counsellor Hand-off
Qualified enquiries and out-of-scope questions escalate to a counsellor with the transcript and a summary, so no student has to repeat themselves.
Four phases
zero shortcuts
Knowledge Mapping
Collected the consultancy's approved material on destinations, courses and entry requirements and structured it into a retrievable knowledge base.
Conversation Design
Defined intents, qualification questions and guardrails: what the assistant answers, what it must escalate, and how it asks for consent.
Model Integration
Connected the LLM to retrieval and the enquiry schema, and tested answers against real student questions before anything went live.
Go-live & Hand-off
Embedded the assistant across the website and wired qualified conversations into the counsellor hand-off flow.
The work,
in full resolution
Drag to explore. Click to open full screen.
Before vs.
after
Algon built us a strong website with brand presence, AI chatbot support, contact flows, and everything we needed to connect better with students.
Want results like these? Let's talk about what we can build together.
Start a project