AI that does the paperwork, not the coaching
The copilot drafts the reviews, reports, and reminders your academy sends every week. It works under four hard rules, and you hold the send button.
Four jobs, four owners
Most AI products blur who does what. ChessCore splits it cleanly, and the boundaries are enforced in code.
Stockfish
Judges every position
All chess evaluation comes from the engine: mistakes, eval swings, missed tactics. The language model never judges a move.
Your database
Supplies every number
Attendance, ratings, fees, and dates are pulled from your records at draft time. The AI cannot invent or alter a figure.
The copilot
Writes the prose
It turns engine lines and database facts into plain language a parent understands, citing the exact move and the eval change.
You
Approve every send
Drafts wait in one queue. Until a coach or admin signs off, nothing reaches a family. Your name is on the approval.
Hours of writing, reduced to a read and a click
Everything below lands in one queue, already written, already sourced from your records. You read, edit if you want, and approve.
Game reviews
A plain-language review of every analyzed game, built only from what Stockfish found.
Progress reports
Monthly per-student narratives wrapped around real attendance, rating, and homework data.
Fee reminders
Polite, specific reminders with the invoice amount and a payment link attached.
Class recaps
A short WhatsApp-ready summary after class: what was covered, what to practice.
AI Copilot queue
3 draftsGame review · Aarav vs. Diya
Stockfish found 2 coachable mistakes · cites 14...Qe6?
March progress report · Batch B2
Drafted for 14 parents · numbers from attendance + ratings
Fee reminder · Sharma family
Razorpay link attached · waits for your sign-off
Auto-send earned on 2 of 9 templates · everything else waits for you
Auto-send is a privilege the AI has to earn
On day one, every template drafts only. A routine class reminder can earn the right to send itself after its drafts prove reliable over logged evaluations.
Anything that touches money, sales, or a minor's guardian never auto-sends. Those messages always wait for a human, no matter how good the drafts get.
And every step is on the record: the draft, the redaction, the approval, the send. If a parent ever asks where a sentence came from, you can show them.
Audit trail · Today
Every action attributableFee updated · ₹4,200 → ₹4,500 · Batch B2
by Admin Meera
March report approved · sent to 14 parents
by Coach Priya
Report drafted · PII redacted before prompt
by AI Copilot
Guardian consent renewed · 2 students
by Sharma family
Built so the worst case cannot happen
Guardrails are not a settings page you have to find. They are how the system works, on every plan, for every academy.
PII is redacted before any prompt
Names, contacts, and identifying details are stripped before text reaches a model. A policy layer ChessCore owns sits in front of every AI call.
Retrieval is tenant-scoped
When the copilot looks something up, the query itself is restricted to your academy's data. Isolation is enforced at the database, not the UI.
Malformed output fails closed
Every generation is checked against a strict schema. If the model returns something unexpected, the draft is discarded, never shown.
Allowances cap every account
AI usage runs within per-plan daily allowances, so a runaway job can never surprise you on cost or volume.
Every AI action is logged
Drafts, approvals, edits, and sends land in the audit trail with who did what and when. Nothing happens off the record.
Auto-send is earned, never assumed
A template can only send without review after its drafts prove reliable over logged evaluations, and you can revoke it any time.
From the blog
AI in chess coaching, explained
Start building a smarter academy
Early access — we set up your academy with you: roster import, Lichess and Chess.com connections, batches, and fee plans.