AI inside Giiki
AI where it earns its keep.
Six capabilities ship inside the modules — not a chatbot bolted on the side. Three are live today. Here is what each one actually does, and the rules they all run under.
The rules every capability runs under
Every AI call is counted per school and billed as used, directly to you, at provider rates — the same pass-through as WhatsApp, SMS and email. You see the meter; nothing hides in a bundle.
Librarians review questions before children see them. Support escalates to people. Verification queues have owners. The AI drafts; your staff decide.
Call analysis runs on-premise — audio never leaves the school’s control. And one school’s data never trains a model for another’s benefit.
Support desk, answered first by AI
LiveThe product's front door is an assistant that already knows the product.
- Every support ticket gets an instant first response, drafted from the product's own context — not a generic chatbot guessing.
- A conversation that needs a human becomes a ticket without retyping — the thread carries over.
- Humans take over on escalation. The AI never closes a ticket on its own.
Reading comprehension
LiveQuizzes for any book in the catalogue, graded like a teacher grades.
- Generates quiz-question drafts for any book — and the librarian reviews before publishing. Nothing reaches a child unreviewed.
- Grades students' written answers against a rubric — real comprehension, not just multiple choice.
- Retake locks and a teacher review queue keep the standards the school's, not the model's.
Recruiter call intelligence
LiveEvery hiring call, understood — without the audio ever leaving campus.
- Every recruitment call is transcribed and attached to the candidate's timeline.
- Analysed for sentiment, interest and call quality — recruiters see the pattern, not just the recording.
- Processed on-premise: audio and transcripts stay inside the school's own environment. DPDP-friendly by design.
AI-verified attendance
Launching · 1 OctA second pair of eyes on every day's roll call.
- A central verification queue where an AI check — or a supervisor — confirms each day's sessions.
- Flags gate-vs-classroom disagreements: a child marked present who never entered the campus becomes a safeguarding follow-up, not a statistic.
- Verification is a queue with an owner, not a silent batch job.
Composable dashboards
Launching · 1 OctDashboards described as a spec, rendered live.
- A dashboard is a description, not a build ticket — described as a spec and rendered on the spot.
- The groundwork for the real goal: asking for any report in plain language and getting it.
Agents for the busywork
PlannedThe operating model where routine admin stops consuming people.
- AI agents take on routine administrative work across the existing assistant and integration seams.
- Read-mostly at first — agents summarise and prepare before they ever act.
- Planned means planned: no date, and we will not promise one.