How to Choose an AI Agent for Student Recruitment: AU Checklist

Use one grid for every vendor, then test two on your own pages

To choose an AI agent for student recruitment in Australia, define eight criteria and weights before you see a demo, score each vendor against them, then run the leading two through a thirty-day pilot on your own course pages. This keeps the decision about your applicants rather than the vendor's sales script.

The guidance below suits universities, TAFE institutes and private higher education providers. It covers the grid, a scoring table, red flags, a pilot with test questions and the Australian regulatory setting. For the basics of what these tools do, see the guide to AI chatbots for student recruitment. When you are ready to put questions to vendors in writing, the chatbot RFP checklist for higher education is the companion document.

What Australian recruitment teams ask of an agent

The Australian pattern combines several pathways: school leavers applying with an ATAR through a state admissions centre, mature-age and VET-to-degree applicants, direct applications to a provider and a large international cohort. Each has different requirements, so the agent must tell them apart early and answer from the right content.

Questions come in predictable waves around offer rounds, orientation and the start of each semester. They also arrive outside Australian business hours, because international applicants live in other time zones. An agent that answers from approved content, books a campus tour or an advising call and flags the right cases for staff relieves the pressure at the busiest points without taking decisions that belong to your team.

Criterion one: answers drawn only from your content

The agent should answer from the course pages and policies you approve, show the source and say "I don't know" when the content is silent. Make every vendor demonstrate this on your site. Try one question that your pages answer, one they do not, and one that tempts a promise: "If my ATAR is 78, will I get into this degree?" The right behaviour is to point to the published entry requirements and any alternative pathways, then offer a conversation with an adviser, without predicting an outcome.

Criterion two: qualification and handoff

The agent should ask a few natural questions, such as study level, domestic or international status, intended start and area of interest, then route to the correct team with a summary. Agree triggers before launch: a request for a person, a complaint, an accessibility request, a financial hardship concern, a welfare disclosure. Ask what the student sees and what the staff member receives at the moment of handoff, and what happens out of hours.

Fee questions need extra care. The agent can explain what you publish about fees and about HECS-HELP loans for eligible students, but it must not assess an individual's eligibility or promise a loan outcome. That goes to a person.

Criterion three: integration, shown rather than described

Ask each vendor to demonstrate how a conversation becomes a CRM record and how an open day, tour or advising appointment reaches your calendar. Provide the same short brief to all vendors: your CRM, your event tool and your consent practice. Then ask which fields are written, how duplicates are handled and who is alerted if a write fails.

Criteria four to eight in brief

Data protection and hosting. Ask where data is stored, who the sub-processors are, whether your content and conversations are used to train models and how long records are kept. Ask for written answers, and ask whether any offshore disclosure occurs.

Multilingual and after-hours. Test the languages of your priority international markets, and test a late-night question.

Review and analytics. Staff need to read transcripts, see unanswered questions and spot content gaps.

Channels. Begin with the website. Treat messaging apps as a separate decision with consent rules of their own.

Pilot, cost and exit. Ask for a time-boxed pilot, an itemised explanation of total cost and a written answer on exporting your data.

The scoring table

Score each criterion from 0 to 5, multiply by the weight and add up. The weights are a suggestion; adjust them and lock them before the first demo.

Use three scorers from different teams, such as admissions, international and IT. Treat a failure on data protection as disqualifying, whatever the total.

Red flags

Pause the process if you see:

• a vendor unwilling to run the agent on your pages before you sign;
• confident answers with no source and no admission of gaps;
• predictions about offers, scholarships or visa outcomes;
• integrations described in the abstract but never shown;
• no written statement of where data is stored;
• your content or conversations being used to train models;
• no pilot, no export right and no exit clause;
• a price tied to a measure that nobody can explain.

The regulatory setting: TEQSA, the Privacy Act and the Go8 context

TEQSA is the national regulator for higher education providers, and it publishes expectations on admissions transparency. What an agent says about entry requirements, pathways and outcomes should be as accurate as your website, because applicants will treat it as your word. The same applies whether you are a member of the Group of Eight or a smaller private provider.

For privacy, the Australian Privacy Principles under the Privacy Act, overseen by the OAIC, set expectations for collecting, using and disclosing personal information, including when information goes to an overseas recipient. Your privacy officer should review the vendor's terms against them. Public universities may also have state-level obligations. For international students there are additional provider obligations, so the agent should point to your published guidance and send individual cases to your international office.

Tell applicants at the outset that they are speaking with an automated assistant.

A thirty-day pilot with test questions

Choose one faculty or one TAFE school. State your success measure first, for example fewer overnight enquiries left waiting and more complete CRM records.

1. Week one. Load approved content, set triggers and connect the CRM and calendar.
2. Week two. Advisers test with the questions below. Fix gaps.
3. Weeks three and four. Go live on selected pages, read transcripts weekly and log handoffs, bookings and unanswered questions.

Test questions:

• "What is the minimum ATAR for this degree, and are there other ways in?"
• "I finished a Certificate IV at TAFE. Can I get credit?" (answer from content or hand off)
• "When do applications close for semester two?" (must match your dates)
• "Can I book a campus tour for Saturday?" (booking flow)
• "Will I definitely get an offer?" (must refuse)
• "Do I pay anything upfront?" (publish what you say, then hand off for individual cases)
• A question in an overseas market's language, sent at 3 a.m. Sydney time.

Score the pilot with the same grid. The comparison of AI chatbots for higher education and the comparison of AI agents for student recruitment can help with the shortlist.

For reference, Skolbot Chat is a website agent that answers from your institution's own content, qualifies applicants, books open days and appointments and pushes the record to your CRM. WhatsApp is an optional module. It deserves the same scrutiny as any vendor.