A Singapore tutoring centre was running its operations on an incumbent scheduling system plus about 15 spreadsheets covering enrolments, scheduling, finance and communications. Qlabs replaced them with one internal platform. Non-technical staff run day-to-day operations through a permissioned chat agent that books trials, records attendance and processes invoices, and it always asks for confirmation before changing anything. The engagement is under NDA, so the centre is not named.
At a glance
| Before | After | |
|---|---|---|
| Systems | An incumbent scheduling system plus about 15 spreadsheets | One internal web platform |
| Areas covered | Split across tools: enrolments, scheduling, finance, communications | One place: students, scheduling, attendance, finance, tutors, marketing |
| How staff work | Switching between the system and spreadsheets | Asking a staff chat agent in plain language, or using the web app |
| Changes to records | Manual edits in several places | Agent proposes, staff confirm, before any write |
| Access | Shared files | Role-based access: admin, finance, marketing, tutor, read-only |
The problem
Many tuition centres start with a scheduling or class-management product and then fill the gaps with spreadsheets: one for trial bookings, one for invoices, one for tutor allocation, one for marketing leads, and so on. Each spreadsheet makes sense on its own. Together they mean the same student appears in several places, staff copy data by hand between them, and only one or two people know how it all fits together. This centre had reached about 15 spreadsheets alongside its scheduling system.
What was built
One internal platform that holds the centre's operations in a single database, with a staff chat agent on top. The agent has more than 40 tools across 10 areas of the business, including students, scheduling, finance, attendance, tutors and marketing. Staff can ask it to do routine work in plain language, such as booking a trial lesson, recording attendance or processing an invoice, instead of hunting through screens and sheets.
Scheduled background workflows handle routine jobs, and the web app is still there for anyone who prefers screens to chat.
Design principles
Confirmation first
The agent clarifies when a request is ambiguous and asks before any write operation. Nothing is booked, recorded or invoiced until a person confirms it. In a business that handles parents' money and children's schedules, this matters more than speed.
Role-based access, checked on every action
Each staff member has a role (admin, finance, marketing, tutor or read-only), and the agent's tools are limited by that role on every call. A tutor can record attendance but cannot issue invoices.
Staff can teach it the centre's rules
Staff can give the agent standing rules for how the centre works. These are stored and applied in later sessions, so the system follows house practice instead of generic defaults.
Choice of where AI runs
The language model runs behind a single interface that supports several backends, including a local model on hardware the business controls. That lets the centre decide where processing of student and parent data happens. Our PDPA and AI checklist explains why that choice matters.
What other centres can take from it
- Count your spreadsheets. If staff keep side spreadsheets next to your main system, those are the places where data gets copied by hand and errors creep in.
- Start with the tasks staff do daily. Trial bookings, attendance and invoices are repetitive, rule-based and easy to check, which makes them good first candidates for AI help.
- Keep a person on every change. Let AI prepare the action; let staff approve it.
- Not every centre needs a custom platform. For many, training staff to use AI tools well on their existing systems is the better first step.
Where Qlabs fits
For most education businesses, the right starting point is our hands-on AI training, which sets AI tools up with the centre's own templates: see AI training for tuition and enrichment centres and for preschools. The course is delivered with an SSG-approved training partner, so eligible employers can use the SkillsFuture Enterprise Credit (what changes on 1 December 2026). When a centre's operations need rebuilding, as this one did, that is a separate consultancy engagement.
Built by Matthew Lee, who has run an education business in Singapore since 2009 and is a Claude Certified Architect (Professional). Email hello@qlabshq.com to talk about your centre.
Frequently asked questions
Which tuition centre is this?
The client is a Singapore tutoring centre. The engagement is under NDA, so we don't name the centre.
Could a centre do this with off-the-shelf software instead?
Often, yes. Off-the-shelf tuition management software suits many centres. A custom platform makes sense when a centre's way of working has outgrown its tools and the gaps are being filled with spreadsheets, as they were here.
Can the AI agent change records on its own?
No. The agent asks before any write operation, such as booking a trial, recording attendance or processing an invoice, and staff confirm it. What each person can do is limited by their role.
Does student and parent data go to an AI company?
The platform can run its language model through different backends, including a local model on hardware the business controls, so the centre can choose where processing happens.
How is this different from the Qlabs AI training?
The training is a short course plus setup of AI tools on the centre's own templates, claimable with the SkillsFuture Enterprise Credit for eligible employers. A platform like this one is a separate consultancy build for centres that need their operations rebuilt.
Do you publish time or cost savings?
Not for this engagement. We only publish figures that were measured against a baseline and that the client has agreed to share.