Your school already has the data. Can you ask

It is a Tuesday in August. The management committee meets at 6pm and the chairman wants to know why collection is behind last year. The principal calls the accountant, who is already home. The accountant promises a report by morning. By morning the question has changed, so the report is the wrong one. This happens in almost every school in India, every month — not because the data is missing, but because getting an answer requires a specific person, at a specific desk, with a specific screen open. The school is sitting on the number. Nobody can reach it in time to use it.

AI is about to change which questions a school can afford to ask — but only if the ERP lets an assistant reach the actual records. A chatbot that explains how to use the software is not the same product as an assistant that reads your ledger. Most vendors are selling the first and calling it the second. That distinction is the whole thing, and it is easy to test.

The useful version of AI access is unglamorous: the school gives an assistant like ChatGPT or Claude permission to read specific parts of its own records, and staff then ask questions in ordinary language. No dashboard, no export, no training. The answer comes from live data, and it can be checked. Ask the same question three ways and it holds. What separates a real implementation from a demo trick is whether the assistant can reach the full record system or just one corner of it, and whether the school controls what it may see.

The India-specific bar is higher than the global one, in two ways. First, staff turnover in school offices is high and formal training rarely survives it, so any system that depends on a clerk remembering which of forty screens holds the answer will quietly stop being used. Asking a question in plain Hindi survives turnover; a training session does not. Second, the data-protection rules that now apply to Indian schools treat student records as sensitive personal data, which means broad, unaudited access is a real liability and not a theoretical one. Any AI access worth buying must narrow access rather than widen it — and must be able to prove, later, exactly what was reached.

How to test it in a demo — five questions

1. "Ask it something only my school would know." Give the salesperson a real question about a real class, live on the call. If the answer needs a report, a filter, or a follow-up email, it is a chatbot, not access. Do not accept a pre-baked demo dataset.

2. "Now ask it a question you didn't prepare for." The second question exposes the first. Prepared demos handle one path well. Ask a follow-up that crosses modules — "of those defaulters, how many also have attendance below 70%?" — and see whether it holds.

3. "Show me what happens when it tries to change something." Ask the vendor to have the assistant modify a record while you watch. A safe system shows a preview and waits. A system that just does it is a system that will one day just do it by mistake.

4. "Show me the permission screen and the log." Ask who decides what the assistant can see, and where the record of its activity lives. If the answer involves emailing support, the school does not control access — the vendor does.

5. "Can I use it from my own ChatGPT or Claude account?" This is the sharpest question of the five. If access only works inside the vendor's own chat box, you are locked into their assistant, their model and their roadmap. Open access means the school picks the assistant.

For context on the market: the names an Indian school will run into while shopping include Teachmint, Entab CampusCare, Fedena, MyClassboard, Vidyalaya, Campus 365 and Edunext. All of them are established products with real deployments, and several have shipped AI features over the past two years. At the time of writing, most of those features are assistants that live inside the vendor's own product — useful, but a different thing from letting the school connect an outside assistant to its own records under its own permissions. That gap is worth checking directly rather than taking anyone's word for, including this page's. Ask each vendor question five and compare the answers.

On pricing, be careful about what AI access does to a quote. Indian school ERPs typically land between roughly ₹50,000 and ₹5 lakh per school per year depending on student count, modules and board, with per-student pricing common at the lower end. The risk with AI is that it arrives as a new line item — an "AI add-on" priced per user or per query, on top of the base licence. Ask early and explicitly whether AI access is included in the annual fee or billed separately, and whether the price changes with usage. A per-query charge quietly discourages the exact behaviour that makes the feature valuable, which is asking freely.

Inkwelly built this as core infrastructure rather than an add-on. Its hosted server exposes 228 tools across 18 modules — fees, attendance, academics, timetable, homework, transport, staff and group-level rollups — so an assistant can reach the same records the dashboard shows. 180 of those tools cannot write anything at all, write permissions sit in a separate grant so read access never becomes write access by accident, and 30 write tools refuse to act until a human confirms a preview. It works from a school's own ChatGPT or Claude account rather than a proprietary chat box, and access is issued and revoked by the school. The step-by-step setup is written up in the connection guide, and the full breakdown lives on the AI access page.

“The question is not whether your school software has AI. It is whether you can ask your school a question and get a real answer — from your own records, in your own words, on your own assistant.”

Decide this in two weeks, not two quarters. Shortlist three vendors, run the same five questions on each demo with a real question from your own school, and write down the answers. The differences will be obvious and they will not be subtle — one vendor will show you a live number, another will show you a chatbot explaining a menu. Then ask the exit question. A school that runs this test properly will know more about all three products in a fortnight than most buyers learn in a year.

Ask it a question about your own school