Proposal · Mastery School of Music · May 2026

An AI-first system for music schools,
live in 10 weeks.

A 24/7 AI concierge that answers parents in any language they search in — English, Mandarin, Cantonese, French, Japanese — even at 11pm. An online enrollment + payments flow on Stripe and Apple Pay so families never come to the front desk to hand over cash. Three portals — parent, teacher, owner. And an AI Ops layer behind it all, doing the boring work seven days a week. Found on Google, recommended by ChatGPT, run by AI.

8–10 weeks  build Staged & refundable  terms Month 1 free  · 50% off months 2–6 24/7 AI Ops  · same system DeployGrow runs on
01 · The friction

What this proposal is actually solving.

Mastery is not short on students. Mastery is short on visibility, automation, and time. Five layers of daily friction quietly cost the school somewhere between $40K and $140K a year — the exact number we calibrate together in week 1.

01

The only inbound channels are phone and email — and neither is monitored 24/7

No web chat. No 24/7 reception. No Instagram DMs. No online booking. A parent searching at 10pm gets nothing back until the front desk picks up the phone the next afternoon — and by then half of them have already booked elsewhere. (Mandarin-speaking families increasingly expect WeChat too — if that audience matters, we'll scope a private-domain solution together in week 1.)

02

Trial classes only happen if the parent insists — and then nothing follows up

The trial flow has no booking page, no automatic confirmation, no pre-class reminder, no post-class “how was it?” nudge. If the parent doesn't chase, the trial doesn't convert. Most don't chase.

03

Parents don't know any of the things that decide whether they enrol

Where do I register? Which teachers can my kid pick? What are their backgrounds? What do other students say? What does it cost? What's the trial price (and why is it different from the website price)? What will my kid actually learn? When does the term start and end? What's the schedule? How much have I already paid? Which order was that? Where is my contract?

  • These are just the surface — every prospective family arrives with their own list. Some answers do exist on the current site, but parents can't find them. Most never get answered at all.
04

The owner is doing the work the system should be doing

There's no central teacher schedule across the seven studios, no instrument / room / timeslot inventory, no source attribution on leads, no weekly numbers. Today the owner holds all of it — schedule, lead history, pricing exceptions, makeup credits — in one head. That's a sign of how well the school is run, not how badly. It's also the single biggest constraint on growth: it doesn't scale past one person, and it never gets a day off.

05

The brand under-sells what's delivered, and there is no acquisition asset that compounds

Seven studios, ten faculty, RCM exam prep, summer camps, recitals — none of it is visible to a parent comparing schools at 10pm on a phone. There is no SEO flywheel, no AI-search optimization (ChatGPT, Google AI Overviews), no nurture sequence, no owned audience. Every new student costs roughly the same to acquire as the last one. Nothing gets cheaper over time.

02 · The system

What gets installed in 8–10 weeks.

One unified system, six layers, all running in production from week 7 onward. You log into one dashboard. We handle everything else.

Layer 01

Lead-generating site

Bilingual EN/ZH front-end, mobile-first, conversion-optimized. Built-in chat, online booking, trial-class flow. RCM / camp / recital pages indexed for both Google and ChatGPT.

Layer 02

Three portals

Parent / Teacher / Owner. Schedule, attendance, makeup credits, contracts, payments, payroll — all inside one login.

Layer 03

Online payments

Stripe + Square integration. Tuition, trial fees, deposits — paid online, receipted automatically. No more cash at the front desk, no more manual POS reconciliation.

Layer 04

Service-email matrix

AWS SES at every step — receipts, pre-class reminders, post-class follow-ups, refund confirmations, trial confirmations, renewal nudges. Every parent touch is on-brand and logged.

Layer 05

9 AI agents · 18 workflows

24/7 multilingual AI concierge on the site (answers in whatever language the parent writes in). Behind every form, message, and event — capture, score, route, follow up, escalate. The owner sees results, not the plumbing.

Layer 06

One owner dashboard

Leads, pipeline, AI activity, financials, alerts, weekly digest — in one screen. Zero new tools to learn.

Click through the live prototype to see all six layers running today → schooldemo.pages.dev/growth-prototype

03 · The unfair advantage

Most agencies sell a website. We sell the AI workforce that runs behind it — the same one DeployGrow runs on every day.

This is the layer that makes a small operation feel like a 20-person one — and it's the layer most proposals never show you. Here it is, fully built, running on a live prototype you can open right now.

AI Ops layer · live in production

9 agents. 18 workflows. 24/7 multilingual concierge. One dashboard.

Every form submission, every chat, every booking, every payment, every email reply — captured, scored, routed, followed up, and emailed back automatically. The owner sees the outcomes; the system handles the rest.

24/7Coverage · zero new hires
9AI agents · always on
18Workflows orchestrated
Any lang.EN / ZH / FR / JP / etc.
24/7 multilingual AI concierge — answers prospect questions on the site at 2am in whatever language they write in (EN, ZH, FR, JP, ES…), qualifies, books trial classes, hands off to humans only when needed
Lead research agent — enriches every new prospect overnight before the owner's morning coffee
Email-reply classifier — reads every inbound email, routes to the right workflow, drafts responses
Payment + receipt agent — handles Stripe webhooks, sends branded receipts, reconciles tuition vs. trial vs. deposit
Service-email orchestrator — pre-class / post-class / refund / renewal / trial-confirmation emails through AWS SES, on-brand, logged, reportable
Refund & makeup-credit agent — applies the 24h / 14-day rules consistently, escalates edge cases
Schedule-optimizer agent — watches teacher availability, room utilization, recommends moves
Reminder & recovery agent — lesson reminders, no-show recovery, re-engagement flows
Reporting agent — weekly performance digest in the owner's inbox — KPIs, alerts, anomalies
+ orchestration layer — 18 workflows wiring it all together (n8n + FastAPI + MCP)

This is a real, running prototype — not a mockup. Click anything; everything works. DeployGrow runs the same agents on its own ops, every day.

04 · We dogfood it

Don't take our word for it. Watch us run on it.

Every agent and workflow we install for Mastery, DeployGrow already runs on itself — every day, in production. We don't sell what we don't already use.

MGAC AI · sister product · InsurTech

100% extraction accuracy on 3,750 ACORD insurance forms in production.

100%

A three-layer pipeline — coordinate-based extraction, structured-field validation, and a vision-model fallback for scanned docs. Zero LLM hallucination on structured fields. The same engineering rigor goes into your school's automations.

DeployGrow · our own ops layer

DeployGrow runs on this exact system, 24/7.

9agents18workflows

Lead research, AI website audits before sales calls, automated payment + receipt emails, post-call follow-up, dashboard reporting, weekly digest. The system Mastery is buying is the system that already runs our company.

05 · Who you're working with

The company. The address. The two people on every call.

No account managers, no junior hand-offs. You get the founders who built this, working on Mastery directly.

DeployGrow

AI business systems for SMBs · Vancouver, BC, Canada

AddressVancouver, BC · Canada
Co-founder & CEO · DeployGrow

Benjamin (Yuze) Gao

10+ years operator. Building DeployGrow's AI ops platform; also co-founder of MGAC AI (Canadian InsurTech).

  • Co-founder, MGAC AI — 100% accuracy on 3,750 ACORD forms in production
  • 10+ years running an enterprise services business in Beijing
  • MS Computer Science, Fairleigh Dickinson University (in progress)
  • MBA, University of Chinese Academy of Social Sciences
  • Mastery parent. I see the front desk every week — this isn't a generic SMB pitch.
Co-founder & CTO · DeployGrow

DAZHI (Alex) Yang

Full-stack + AI/ML + big-data engineer. Anthropic Claude Code certified. Builds the systems we ship.

  • Principal engineer (prior) — research stack with FDU / SFU / BCIT, 99.9% uptime, 40+ live student users
  • Independently built TalkCub — full online education platform (user app + backend + admin)
  • Big-data engineer (prior) — Sichuan telco signaling, 90M users / day, 400M+ records / day
  • React 18 · TypeScript · Spring Boot 3 · MongoDB · Neo4j · OpenAI / DeepSeek / Anthropic
  • MS Computer Science, Fairleigh Dickinson University (in progress)
  • Anthropic Claude Code certified · active robotics RL practitioner

DeployGrow is a young company. Mastery would be our flagship Vancouver client. We're under-pricing, over-delivering, and staging payments because that's how trust gets built when there isn't a long client list yet. We'd rather earn the case study than sell the contract.

06 · Year-2 and beyond

Year 1 you pay for the system. Year 2+ the system pays you back.

Most agencies finish at “site live, dashboards green.” That's where the compounding actually starts. Four flywheels turn on around month 6 and keep turning.

Wheel 01

Local SEO that compounds

Every recital recap, every faculty interview, every program page becomes a long-tail asset. Acquisition cost drops every month it runs.

Wheel 02

Owned-audience nurture

Email + SMS + WeChat sequences via AWS SES + the service-email orchestrator. Klaviyo's published benchmarks show automated email flows generate roughly 30× the revenue per recipient of one-off campaigns. We build the flows; the audience compounds.

Wheel 03

WeChat private domain

For Mandarin-speaking Vancouver families — group + 1:1 + broadcast. Western agencies can't service this; we do, in language.

Wheel 04

AI-search optimization (GEO)

Parents are now searching ChatGPT and Google AI Overviews for “piano lessons UBC area.” We optimize the school for AI retrieval before competitors realize this is a channel.

Each wheel reduces dependence on paid ads, and each one strengthens the others.
07 · Investment

Staged. Refundable. First month free.

$1,000 to start, three refundable milestones, and a sliding subscription that protects you while we earn the work. If you fire us at any milestone, the refundable portion comes back.

Setup · 8–10 weeks · staged

$9,800 build, paid in 4 stages

$1,000 locked + $8,800 across 3 refundable milestones
Discovery kickoff
Day 0 · to start
$1,000
Milestone 1 · Foundation
Week 3 · site + portals + payments
$3,000
Milestone 2 · AI Ops live
Week 7 · 9 agents + 18 flows + SES
$3,000
Milestone 3 · Polish & launch
Week 10 · go-live
$2,800
Setup total$9,800
Subscription · first year discount. Month 1 free. Months 2–6 at $1,250 / mo (50% off). Month 7+ at $2,500 / mo. Hosting, monitoring, agent updates, AWS SES, and ongoing optimization included.
First-year total

$31,050

Saves $8,750

vs. flat $9,800 setup + 12×$2,500/mo = $39,800. The discount is real, not a number trick.

Risk reversal

If we miss a milestone, you don't pay it.

Each milestone has a written acceptance gate — functional, demo-able, your sign-off. If we fail to ship, that milestone payment is refunded. Discovery is the only locked stage; everything else is yours to claw back.

Variable usage costs — LLM API calls, SES email volume, SMS, Stripe fees — pass-through at vendor cost. Typical music school usage runs $80–$220 / month and is fully visible in the dashboard.

08 · Phase-2 upgrade · ready when you are

Two to three months after Growth is humming, add native mobile apps.

Growth ships a mobile-responsive web that already works on every phone. Scale upgrades the parent / teacher / owner experience to true native iOS & Android apps — with deeper hooks into the phone itself. Same system underneath, more places it can reach the user.

Scale package · phase 2

What native could unlock for Mastery

Phase-2 scope gets shaped by what you actually need once Growth is live. A few examples of moments that only work when there's a real app in the parent's, teacher's, or owner's pocket:

  • Lock-screen lesson reminders. Parents see “Aaron's piano lesson in 30 min · Studio 4” without unlocking the phone.
  • One-tap arrival check-in. Front desk stops chasing names; the teacher sees who's here in real time.
  • Push approvals for the owner. Makeup credits, refunds, schedule swaps — decided on the phone in seconds, not 12 emails.
  • Native payment & offline-first attendance. Apple Pay / Google Pay tuition on the parent app, plus attendance that works in the elevator or the studio basement.

These are starting points, not a fixed feature list. We co-design the real scope with you once we've watched Growth run for 2–3 months and we know what your day actually looks like.

Setup$18,000vs $9,800 Growth · adds full iOS + Android build, store submission, push infrastructure
Monthly$4,500 / movs $2,500 Growth · adds app maintenance, store updates, native push delivery
TimingPhase 2Recommended after Growth has been live 2–3 months and the workflows are calibrated
Open the mobile preview →

Phase-2 mobile is opt-in. Most clients run on Growth-only for 6+ months before upgrading.

09 · ROI · with sources

What the system has to do to pay for itself.

Our base case assumes only +10% trial-to-paid conversion uplift. The published industry benchmarks below report 12–40% — we've deliberately set ours below the floor so the math survives a bad month.

Mastery base case (conservative)

+10% conversion uplift

~$42,000

Recovered first-year revenue, assuming 70 trials/mo · $3,500 student LTV · +10% trial-to-paid lift. System pays itself back in month 8–9.

If results land at industry median

+25% conversion uplift

~$105,000

Same school, same students, mid-range industry-benchmark uplift. The exact number we calibrate together using your actual baseline in week 1.

Published benchmarks behind those numbers — every source clickable

Our +10% base case sits deliberately below the 12–40% range these studies report. If we hit only the floor, the system pays itself back. If we hit the median, you get a flywheel.

10 · Two ways to start

Pay $1,000 today, or talk to us first.

Either path works. Discovery is the only locked stage — everything else is refundable.

Path A · Recommended

Pay $1,000 & start discovery

We kick off this week. Week 1 = baseline calibration + scope sign-off. If we don't agree on scope by end of week 1, you walk — the $1,000 is the only locked fee.

Pay $1,000 · Start now →

Stripe checkout. Receipt + welcome email + week-1 calendar within 5 minutes.

Path B · If you want to talk first

Book a 20-min call

Both founders on the call. We open the prototype together, walk through your numbers, answer everything. No pitch deck.

Book 20 min →
DeployGrow

AI business systems for SMBs · Vancouver, BC, Canada

AddressVancouver, BC · Canada

Benjamin Gao · CEO

WeChatgbenjamin

DAZHI (Alex) Yang · CTO