The Real Problem
Priya is the homestay coordinator at a language school in Christchurch. She runs a network of 80 host families across the city. She is also the welfare officer, the dispute mediator, the airport pickup organiser, the family-recruiter, the meal-allowance-questioner, and the person every student calls when the wifi at their homestay is too slow.
It's a Tuesday. The phone buzzes at 11:47pm. It's Margaret, a host mum out in Sumner. She's worried about Mei, a 17-year-old Chinese student who arrived nine days ago. Mei has been crying in her room every evening, hardly eating, won't come downstairs for dinner. Margaret has tried to talk to her but Mei's English isn't strong enough yet, and Margaret doesn't speak any Mandarin.
Priya sits up in bed. This is the call she's been afraid of. Under the Education (Pastoral Care of Tertiary and International Learners) Code of Practice 2021, the school is responsible for the wellbeing of every international student under 18. A homestay welfare issue that goes unresolved for 24 hours becomes a Code of Practice incident. An incident that goes unresolved for a week becomes an NZQA complaint. A pattern of incidents becomes a deregistration risk.
Priya emails the school's Mandarin-speaking academic coordinator, who's asleep. She emails Mei's agent in Shanghai, who is in a different time zone and won't reply for hours. She drives to Sumner at 7am the next morning to do a face-to-face check-in.
Mei is fine, eventually. She was overwhelmed, missed her family, didn't know how to ask for the rice cooker she'd been using at home, and didn't realise she could request a different meal plan. Margaret, the host mum, didn't know that asking for cultural food preferences in week one is normal. Two small misunderstandings, one panicked midnight phone call, one school-day-long crisis intervention.
This is one student in week one. Priya has 60 students arriving in the next intake. Her placement spreadsheet is a 23-tab Google Sheet. Her welfare check process is a calendar reminder to phone each student at week 2 and week 6. Her Code of Practice incident log is a Word doc on her desktop. She is one person doing the job of three, and the only thing that scales is the risk.
Why Existing Tools Don't Solve This
SELMA and Class track enrolment, attendance, and academic progress. Neither of them touches accommodation, host family relationships, or pastoral welfare. The data sits in a different system, usually a spreadsheet, sometimes a database the school built in 2014 that nobody understands anymore.
Homestay-specific platforms like StudentHomestay.com or HomestayBay handle the booking and payment side, but they treat the student-family relationship as a transaction, not an ongoing pastoral care responsibility. They don't help you spot a welfare issue forming in week one.
ENZ's Code of Practice templates are useful starting documents but they assume you have time to fill them in. A school running 80 homestay placements on one coordinator does not have time to fill in 80 individualised welfare review forms every term.
The pastoral care coordinator's brain is the most reliable system most schools have. It also doesn't scale, doesn't survive staff turnover, and doesn't pass an NZQA audit because none of the knowledge is written down.
No off-the-shelf NZ tool combines homestay matching, week-by-week welfare check automation, complaint tracking, and Code of Practice incident logging in one place.
How AI Solves This
The school connects its student arrival data, host family profiles, and existing welfare check process to an AI agent. The AI runs the routine work and keeps Priya focused on the conversations only a human should have.
Pre-arrival matching
Before a student arrives, the AI cross-references the student's intake form (dietary needs, religion, age, hobbies, language, allergies, family situation) against host family profiles (existing students in the home, family composition, dietary willingness, location relative to school, previous student feedback). It produces a ranked shortlist of three suitable families. Priya reviews the top match and either confirms or overrides.
What used to take 90 minutes per student becomes a 5-minute review.
Week-one wellness sequence
On day 2 after arrival, the AI sends a wellness check to the student in their first language (Mandarin, Korean, Japanese, Spanish, Portuguese, Thai, Vietnamese, whatever the school's intakes need). The questions are simple:
你昨天晚上睡得好吗？(Did you sleep well last night?) 寄宿家庭的食物你能吃得习惯吗？(Are you comfortable with the food at your homestay?) 有什么想跟我们说的吗？(Is there anything you want to tell us?)
If the student replies in their own language with anything that signals discomfort, such as short replies, no replies, or words like 难过 (sad), 想家 (homesick), 不舒服 (uncomfortable), the AI escalates to Priya immediately with a translation and the conversation transcript. If the student replies positively, the AI logs the check-in and moves on.
The same sequence runs again on day 7, day 14, and day 30, with different questions tuned for each stage of culture-shock recovery.
Host family check-in
In parallel, the AI messages the host family in English on day 3 and day 10. Margaret would have been asked:
Hi Margaret, how is Mei settling in? Is there anything that's confused you about her food preferences, sleeping patterns, or daily routine?
Margaret could have raised "she's been quiet at dinner and goes straight to her room" on day 3 instead of breaking down at midnight on day 9. Priya would have had a Mandarin coordinator on the case during business hours, not been driving to Sumner before sunrise.
Code of Practice incident log
Every welfare check, every escalation, every host family communication, every resolution gets timestamped and logged in a Code of Practice-formatted incident register. When NZQA's quality assurance team asks for evidence at the next external evaluation, the school exports a single PDF that shows every welfare action taken, by whom, on what date, with what outcome.
What used to be a Priya-shaped knowledge-management black hole becomes a documented, auditable, defensible compliance record.
How We Set This Up
None of this works if the AI is just a generic chatbot bolted onto your school. That's why BestAI builds a custom integration program that bridges your AI assistant with the systems your homestay operation actually runs on.
For this kind of setup, that means:
- Connecting the AI to your student management system (SELMA, Class, or a school-built database) so it knows when a student is arriving and where they're being placed
- Connecting it to your host family records (often a spreadsheet or Airtable) so matching can run on real data
- Setting up multilingual messaging through the channel your students actually use, usually WhatsApp, sometimes WeChat, sometimes SMS
- Configuring the welfare check cadence and trigger words for each language
- Building a Code of Practice incident register that exports straight to NZQA's preferred audit format
Our process:
- We map your placement workflow, including the 23-tab spreadsheet, the Word doc incident log, and the calendar reminders that live in Priya's head.
- We build the connections. Our developers write a custom program (an API connector) that lets the AI read your student data and host family data securely.
- We translate the welfare check questions into every language your school commonly takes intake from. A native speaker reviews each set before launch.
- We define the escalation rules: what triggers an immediate call to Priya, what triggers a same-day email, and what just gets logged.
- We test with one intake, usually 10-15 students, comparing AI welfare flags against Priya's manual instinct, before rolling out to the full cohort.
You don't need to be technical. We handle all the development. You just tell us what a "good" homestay placement looks like at your school, and we build the system to find them and protect them.
The Result
- Welfare issues caught in week one, not at midnight in week two
- Homestay matching reduced from 90 minutes to 5 minutes per student, freeing the coordinator to actually talk to students who need help
- Multilingual welfare checks running automatically in the languages your intake actually speaks
- Host families feel supported because someone is asking them how it's going, not waiting for them to escalate
- Code of Practice incident log audit-ready at any time, not reconstructed from emails the night before NZQA arrives
- Coordinator role becomes sustainable, not a slow-burn burnout job that turns over every 18 months
For a school with 200 homestay placements per year, even one prevented incident pays for the system. That could be a student who would have gone home, an NZQA complaint that would have escalated, or a host family that would have quit the program. Most schools that adopt this report 3-5 prevented incidents in the first term alone.
What AI Can't Do Here
- AI doesn't replace Priya. It catches the routine signals so she can spend her time on the hard cases that need a human
- AI doesn't make a 17-year-old feel safe. It tells you, on day 3, that she might not feel safe yet, so a human can have that conversation
- AI doesn't decide whether to remove a student from a host family. It surfaces the warning signs. The pastoral care decision stays with qualified school staff
- AI doesn't substitute for the school's Code of Practice obligations. It documents your compliance with them
- AI won't translate cultural nuance. The ranked shortlist still goes to a human who knows the host family and the student's profile
Who This Is For
- Language schools running 30+ homestay placements per term where one coordinator is doing the work of three
- Schools that take students under 18 (where Code of Practice obligations are strictest)
- Schools that have had an NZQA quality assurance finding around pastoral care documentation
- Coordinators who have been woken up by a host family at midnight at least once
- Any school where the homestay program is the difference between a student finishing the course and going home in week three