Ano ang nakikita ng AI, at ano ang hindi.
Teamer uses AI only when staff ask it to draft a survey or categorise answers to a written-response question. Analytics runs separately, whether or not AI is used.
Ano ang ipinapadala ng bawat gamit, at kailan
| Use | Provider and region | When |
|---|---|---|
| Survey drafting | OpenAI, United States. | Only when a staff member asks Teamer to draft or revise a survey. |
| Response categorisation | OpenAI, United States. | Only when staff generate AI categories for a written-response question. |
| Routine analytics, not AI | PostHog's US cloud, United States. | Whenever anyone uses the site or app. This does not depend on staff choosing to use AI. |
What each use receives
Survey drafting
The staff member's request, course and assignment context such as the course name and term, and, when revising, the survey's current title, description and questions.
Response categorisation
A written-response question, students' answers to it labelled with temporary codes, the course's grouping settings, and category labels, including names typed by the coordinator, used to assign answers. A labels-only safety rename can send sensitive suggested labels back to OpenAI.
Routine analytics, not AI
Product analytics, error reports, diagnostic logs and session replays: site and app usage, device and page information, and user and session identifiers. For anyone signed in, staff or student, their account ID, name, email and platform-owner status. Session replays record what appears on screen, so they can include student names, survey answers and complaint details.
Teamer does not attach a student's name, email address or student number to AI requests. Written text is not checked for personal information. Anything a student types into an answer, or staff type into a prompt or survey, can be sent.
When AI is involved at all
For most of what Teamer does, no AI is involved. Building surveys, collecting responses, managing enrolments, editing teams, and publishing them all run without it. OpenAI is used in two optional, staff-requested workflows: drafting a survey and categorising answers to a written-response question so they can be used in team matching. Neither workflow runs in the background or merely because a course, survey, or response exists.
Survey drafting: what OpenAI receives and why
When a staff member asks Teamer to draft a survey, Teamer sends OpenAI the request they typed, along with basic course and assignment context such as the course name and term. When revising an existing survey, it also sends that survey's current title, description and questions.
OpenAI uses that context to propose a survey title, description, and questions. This workflow does not fetch or attach a student roster, student record, or stored student response. Generating a draft does not publish a survey; staff can review and edit it before students see it.
Response categorisation: what OpenAI receives and why
When staff generate AI categories for a written-response question, Teamer sends the question and the students' answers to it, along with the basic grouping settings for the course. Each answer is labelled with a temporary code, not a student or database identifier. Requests that assign answers to categories also send the category labels, including names typed by the coordinator.
If suggested labels describe sensitive personal circumstances, Teamer can make a labels-only safety rename request to OpenAI. This sends those suggested labels without the question or answers, to replace them with neutral descriptions or Other. The suggested labels themselves may contain sensitive information.
OpenAI returns suggested categories, which Teamer links back to responses internally. Staff can check and correct them. Teamer's own matching rules, not OpenAI, then use the categories as one input when staff generate draft teams; OpenAI does not form or publish teams.
What de-identified means here
Teamer does not select or attach a student's name, email address, student number, Teamer record ID, or roster entry to either type of request. These workflows do not send those student records to OpenAI.
This is not content redaction. Teamer removes formatting and shortens very long answers, but it does not check answers or staff-provided survey content for personal information. If a student includes personal information in an answer, or a staff member includes it in a drafting request or survey text, that information can be part of the request. Staff should not enter personal information into AI drafting prompts or survey text, and should tell students not to put it in answers to a written-response question that will be categorised.
What leaves Australia, and to whom
Teamer's database and file storage are hosted in Australia, in the Sydney region on Supabase, with the app running on Vercel in Sydney. Both run on AWS. The two optional workflows above send the content described to OpenAI's API in the United States. Teamer makes no other use of OpenAI.
Under OpenAI's API terms, data sent through the API is not used to train its models unless the customer opts in. OpenAI may keep it in abuse-monitoring logs for up to 30 days.
Analytics, separate from AI
Separately, PostHog's US cloud processes product analytics, error reports, diagnostic logs and session replays in the United States. It receives site and app usage, device and page information, user and session identifiers, and the account IDs, names, emails and platform-owner status of anyone signed in, staff or student. Replays can include course records visible on screen, such as student names, survey answers and complaint details. This collection does not depend on staff choosing to use AI; turning off the optional AI features does not turn off analytics or session replay.
What stays a human decision
The AI suggests, it does not decide. Its output is a set of suggested categories, not teams. Draft teams come from Teamer's matching rules and land in front of staff, who can inspect them, move students, and change anything before they become teams.
No team reaches a student until a staff member publishes it. If you would rather not use AI at all, skip AI categorisation: written answers then do not affect matching, while draft generation from availability, choices, scales and classmate picks, and building teams by hand, work as before.
What Teamer does not claim
Teamer is an independent product, built and run from Sydney. It is not endorsed or approved by any university, and it holds no SOC 2, ISO 27001, or IRAP certification. Nothing here is meant to imply otherwise.
Nakalista ang bawat provider na ginagamit ng Teamer sa Tiwala at lokasyon ng datos. Para sa iba pang tanong, mag-email sa security@teamer.university.
Magsimula ng pilot