How MONA fully automated 30 businesses with AI, part 1

Founders and mid-market chain owners, we at MONA wrote this post to share something quite rare in Vietnamese online content: full-automation case studies of businesses at real deployment, real operation, real measurement. We have already covered the post AI-First 2026 on strategy. This post is the execution part. Sharing the first 3 representative cases that MONA has business automation for long-standing clients in 2024-2026: an English center chain with ~22 branches, a spa and nail chain with ~28 branches across three regions, and a gold shop system with 12 branches in HCMC. No demos, no theory. The software runs for real, with real interface screenshots and real measured numbers. Read more perspective in 30-day roadmap for CEOs to use AI for a frame of reference. Read more perspective in applying AI in the HR department for a frame of reference. See also how employees react to AI rollouts, as MONA has observed.

Context: McKinsey's State of AI 2025 survey found that businesses deploying production-grade automation see EBIT rise by 8-12% on average in the first 12-18 months. The results MONA measures across 30 long-standing clients sit at the top of that range. That is because we don't sell automation the way slides do. The post the 12-month AI-first roadmap for businesses is recommended further reading. The article why AI projects die early in Vietnamese companies is recommended further reading.

Before you read the 3 cases: how MONA's automation differs from a drag-and-drop consultant

We want you to grasp the core first: MONA does not sell drag-and-drop tools. The Vietnamese market in 2025-2026 is full of "automation consultants" offering to build workflows with n8n, Zapier or Make. Only 5,000,000 to 10,000,000 VND, shipped in 1 week. It sounds cheap. See details in the post lessons from 30 MONA AI deployments.

We understand why you like it. But 6 months later is a different story:

• The cloud tool goes down → the whole automated flow stops. Your business stands still.
• Edge cases (VietQR fails, SMS brandname rejected, invoice numbers out of sync) → can't be debugged because there is no source code.
• Scale volume 10x → the free tier chokes, and the paid plan costs more than one employee's salary.
• Switch providers = rebuild from scratch. Vendor lock-in.

We at MONA choose differently: write real automation software, with source code built to run for 5-10 years, deployed on your own servers and deeply integrated with your core systems. We link module to module in a closed loop: admissions hands off to finance by itself, finance to reporting, reporting to management. No one carries things by hand between steps. MONA has documented this in full in the post AI for customer care. MONA has written this up in full in AI for spa chains. MONA has written this up in full in 5 reasons enterprise AI projects collapse.

2-3 times more expensive at the start. But it runs for years without breaking, scales with the business, and has the MONA team maintaining it long term. MONA's 30 long-standing clients, all mid-market or above. They have run this model for 2-5 years, and none has gone back to drag-and-drop tools.

OK. On to the 3 cases.

Case #1. Automating admissions and grading for a chain of ~22 English centers

Client scale at MONA

A mid-market English center chain ~22 branches in HCMC, Hanoi and Da Nang, ~15,000 students enrolled, revenue around 38,000,000,000 VND/month. The founder has 12 years in the industry.

Opening. Does this scene look familiar?

You know this scene: at 5:01 PM a parent messages on Zalo with one simple question about their child's class. Exactly 1 minute after clock-out. The next morning at 9 AM a staff member finally replies "Sorry, I just started my shift"? By then the parent has signed up with another center. We get it. The founder of this chain put it plainly:

"I lose about 30% of leads every month because a telesales team of 35 can't cover 22 branches. Everyone wants weekends off. A parent asks about the curriculum, the rep doesn't know and has to ask a teacher. The teacher is in class and can't reply. The lead goes cold. A competitor takes it."

The second pain is grading. 28 full-time teachers do almost nothing but two things: grading Writing and grading Speaking. Salary plus allowances of ~22,000,000 VND per person comes to about 616,000,000 VND/month for one repetitive task. Parents still complain, "my child's essay was submitted 2 weeks ago and there's still no feedback."

The third pain is the monthly parent report. A dedicated HR team of 4: one page per student, 15,000 pages a month. Some months it wrapped up close to Tet, and parents vented on Facebook.

What MONA built for this chain

We wrote a dedicated software suite called Mona EduCenter, not a drag-and-drop workflow. It has 3 modules linked into a closed loop:

1. AI SDR admissions module. Integrates Zalo OA, Facebook Messenger, web forms and the hotline. The AI Agent answers 24/7, qualifies leads against 12 criteria, books consultations into staff calendars and sends a reminder 1 day before. Whatever a parent asks gets an answer in 3 seconds.
2. AI Speaking and Writing grading module. Students record themselves speaking English → AI scores to the standard Cambridge IELTS bands and gives detailed feedback. Writing works the same way. It grades Task 1 and Task 2, flags grammar errors and suggests how to raise the band. Teachers only review a random 10% for quality control.
3. Module Auto Parent Report. The software pulls scores, progress and assignments from the LMS, generates a personalized report for each student and sends it to parents by Zalo and email at exactly 7 PM on the last day of the month.

The interface screenshot above is this morning's admissions dashboard. 247 leads in, 189 qualified by AI, 94 consultations booked, 1,247 assignments graded by AI since the start of the day. All running on its own.

Results after 8 months of operation (Q4/2024 → Q2/2025)

• Telesales team cut from 35 to 8 people, saving about 378,000,000 VND/month. The remaining 8 became closers for big deals, on higher pay.
• Grading team cut from 28 teachers to 4 reviewers, saving about 528,000,000 VND/month. The other 24 teachers moved to teaching advanced classes (higher pay, direct revenue).
• The HR team no longer writes parent reports, and the 4 people moved to VIP customer care and community events.
• Enrollment up 47% in 8 months (only 12% in the same period the year before). No more missed leads at night and on weekends.
• Parent satisfaction (NPS) from 42 to 71, mainly thanks to response speed and steady reporting.

Total fixed-cost savings: ~1,000,000,000 VND/month. MONA's build cost over the first 4 months (audit + design + software + integration): about 2,400,000,000 VND. Payback in under 3 months from headcount savings alone.

The same features can also apply to these industries

If you run a business outside English centers, take a look too, because the core pain points are the same:

• Academic tutoring centers (math, physics, chemistry for grades 10-12, university exam prep, IELTS, SAT). AI SDR qualifies parent leads, AI grades assignments against the rubric, and parent reports go out on schedule.
• Bilingual preschool and kindergarten chains, AI admissions advice and automatic weekly child development reports to parents on Zalo.
• Vocational and workplace skills centers, AI SDR qualifies corporate learners and certificates are issued automatically.
• B2C online academies (like Khánh Hùng Academy). Automatic student onboarding, AI grading of practice work and automatic progress reminders.
• Sports and kids' talent center chains, AI booking for trial classes, automatic attendance tracking and parent notifications.

Case #2. Automating booking and customer care for a chain of 28 mid-market spa-nail salons

Client scale at MONA

A mid-market spa and nail chain ~28 branches in HCMC, Hanoi, Da Nang and Can Tho, revenue around 52,000,000,000 VND/month. The founder has 9 years in the beauty industry.

The original pain. The founder's exact words to MONA

We met the founder in October 2024. The first thing she said:

"I don't need a marketing strategy anymore. Customers fill my inbox and my team can't keep up. On weekends Zalo explodes, and on Monday morning I open 800 unread messages. I lose revenue at night. I've measured it."

Digging in, there were 5 pains:

1. 38 Zalo customer care staff split into 3 shifts covering 28 branches. Still too slow. A customer asks if a slot is open, and the reply comes 30 minutes later. They've already booked elsewhere.
2. Messy weekend booking: 2 staff book 1 technician for 2 customers at the same time. The customers show up and wait. Then comes the argument over whose fault it is.
3. 18% no-show rate, far too high. Technicians sit idle and revenue drops.
4. End-of-day revenue reconciliation takes 6 hours, with 3 full-time accountants adding up 28 branches by hand, and a 3-5% error rate.
5. Customer care staff who "clock out": machines off at 6 PM, so customers booking in the evening wait until the next morning.

MONA wrote Mona Spa for this chain

We wrote a dedicated software suite called Mona Spa, a chain of 6 modules linked end to end:

1. 24/7 AI Booking over Zalo and Messenger. The AI Agent reads the customer's need, checks slots in real time across technician calendars at 28 branches, suggests 2-3 fitting options and closes the booking in 38 seconds on average. You can see a sample conversation in the image above.
2. Auto Deposit qua VietQR. AI generates a QR code for a 100,000 VND deposit → customer scans → deposit confirmed → booking auto-confirmed → slot blocked. No-shows dropped to 4%.
3. Auto Reminder. 1 day before at 9 AM → Zalo reminder. 2 hours before → SMS. The technician is notified 30 minutes ahead.
4. Auto check-in and check-out. The customer scans a QR at the counter → service fee, tip and voucher are calculated automatically. The technician doesn't need to press anything.
5. Automatic end-of-day revenue reconciliation. At 9 PM it closes the books for all 28 branches, matches cash, bank transfers and vouchers, and sends the report to the board on Zalo. Error under 0.1%.
6. Auto CSAT survey and rebooking. 24 hours after the service → a short survey, a 5% voucher for the next visit and a suggested next appointment based on the customer's usual frequency.

Results after 6 months (Q1/2025 → Q2/2025)

• Customer care cut from 38 to 8 people, saving about 510,000,000 VND/month. The remaining 8 only handle the hard cases the AI escalates.
• Reconciliation accountants cut from 3 to 1, saving about 50,000,000 VND/month.
• No-shows from 18% down to 4%, which alone brings in an extra ~6,500,000,000 VND revenue/month (technicians are no longer sitting idle).
• Reconciliation from 6 hours/day to 0 hours/day, with the software closing the books at 9 PM.
• After-hours bookings up 340%, with customers booking between 10 PM and 2 AM and the AI still booking them. This is the busy office-worker segment. The best part.
• CSAT from 4.1 to 4.7/5, with customers praising the speed and being remembered by name.

MONA's build cost: about 1,800,000,000 VND. Payback in under 2 months if you also count the revenue saved from no-shows.

The same features can also apply to these industries

If you run a business outside spa and nail, take a look too, because the pain of "booking + reconciliation + customer reminders" is the same:

• Hair salon chains, 24/7 AI booking over Zalo, automatic reminders and automatic stylist commission reconciliation.
• Dental clinic chains, AI service advice, booking by doctor and automatic reminders for implant and braces treatment plans.
• Aesthetic clinic and dermatology chains, AI treatment advice and automatic progress tracking from before-and-after photos.
• Yoga and pilates studio chains, AI class booking, automatic reminders and automatic renewal of the next package.
• Small to mid-size general clinic chains, AI symptom advice, booking the right doctor and automatic follow-up reminders.
• Physical therapy and traditional medicine center chains, automatic booking by treatment plan and automatic tracking of intensity and frequency.

Case #3. Automating transactions and compliance for a system of 12 gold shops in HCMC

Client scale at MONA

A gold shop system 12 branches in HCMC (top 10 gold markets in the South), revenue around 280,000,000,000 VND/month. The owner has 23 years in the industry.

Why are gold shops the hardest of the 3 cases to automate?

We thought gold was hard at first too. Prices move in real time, everything is weighed to the chi, State Bank (SBV) rules are strict, customers want printed paper receipts, and there is also pawn business and gold bar dealing. In 2024 the owner came to MONA with pains that anyone in the same industry will recognize:

• The gold price jumps every 15 minutes. 60+ tellers have to open SJC, open PNJ, write into a ledger by hand and use a pocket calculator, so errors are easy. Customers stand waiting, impatient.
• End-of-day reconciliation takes 5-6 hours and ends in arguments. 12 branches report into a Zalo group, 5 accountants add up by hand, and an error of 1-2 chi is nearly 100,000,000 VND.
• SBV reports done by hand. 6-8 hours every week. Some days they were late and nearly got fined.
• Anti-money laundering (AML): transactions above 300,000,000 VND must be reported. Staff sometimes forgot. The owner lost sleep.
• Fear of staff "skimming" prices. A customer sells to the shop, the staff quote a lower price and pocket the difference. Hard to control with 60 people.

MONA wrote Mona JMS. End-to-end gold shop software

1. Real-time sync of SJC + PNJ + SBV prices every 15 seconds. The software updates itself, so staff don't write anything by hand. You can see the gold price ticker at the top of the screen in the screenshot.
2. Automatic AI price quotes. A customer sells a 9999 ring weighing 2.4 chi → the software calculates: SJC buy price × weight × (100% − 1.2% melting fee) + VIP bonus. A clean number for the customer in 2 seconds. Staff can't "skim" because they don't decide the price.
3. Ledger synced across 12 branches in real time. Every closed transaction goes into the central ledger within 1 second. The owner sees revenue for all 12 branches on one screen.
4. AI Compliance NHNN. Transactions above 300,000,000 VND are flagged automatically, a CTR (Currency Transaction Report) form is generated and sent to the SBV system on time. Anti-money laundering checks run automatically against the blacklist.
5. End-of-day report closes automatically at 9 PM. Reconciles 12 branches, matches gold inventory, estimates gross profit, and sends it to the owner on Zalo.

Results after 10 months (Q3/2024 → Q2/2025)

• Transaction time down 60%, from ~6 minutes to ~2.4 minutes. Customers don't wait.
• End-of-day reconciliation from 5-6 hours down to 0 hours, done by the software at 9 PM. The next morning the owner opens Zalo to read the report.
• SBV reports sent on time 100% of the time. No incidents since Q4/2024.
• Revenue up 28%, because customers don't wait and so don't walk to competitors.
• Gross profit up 4.2 percentage points, because AI quotes more accurately than low-skilled staff and can't skim in reverse. This part alone brings in an extra ~12,000,000,000 VND gross profit/month.

MONA's build cost: about 3,200,000,000 VND. Payback in under 3 months counting only the added gross profit.

The same features can also apply to these industries

If you run a business outside gold, take a look too, because the pain of "real-time price swings + reconciling high-value transactions + compliance" is the same:

• Jewelry store chains (PNJ-style, DOJI-style). AI quotes jewelry and gemstones off the real-time gold price, with a synced ledger.
• Pawn shop chains, AI valuation of pledged assets at market price, automatic tracking of redemption deadlines and automatic SBV reports.
• Foreign exchange and currency exchange dealers, real-time USD/EUR/JPY rate sync and automatic reports on large transactions.
• Luxury watch and gemstone stores, AI valuation against market benchmarks and automatic inventory tracking by serial number.
• Luxury and imported car dealers, automatic reconciliation of high-value transactions, plus automatic compliance and tax filing.
• Online lottery agents and state lottery dealers, automatic multi-branch revenue reconciliation and automatic State Bank compliance.

4 principles MONA always follows when fully automating a business

Before the 4 principles, you need the broader context. Stanford HAI's AI Index 2026 report shows that 78% of businesses worldwide have tried at least one AI use case, but only 22% moved to production-grade. The rest are stuck at the pilot stage. The 30 long-standing clients we work with are exactly the group that made it to production. The four principles below explain why. Read more perspective in AI for sales teams for a frame of reference. Read more perspective in applying AI to the accounting department for a frame of reference.

Reading the 3 cases above, you can see a common pattern. We'll share plainly the 4 principles MONA keeps when taking on a full-automation brief. Use them whether or not you choose to work with MONA:

1. Don't automate the old process. Redesign the process around the new capability. Most consultants drop AI into the old process and proudly say "we have AI now." We do the opposite: look at what AI can do in 2026, redraw the process from zero and throw out what's redundant. The 28-branch spa chain above cut 9 manual customer care steps down to 1 AI booking step. That is redesign.
2. Module links to module, no drag-and-drop tools. Every module MONA writes is its own source code, with tests, logs and monitoring. One module calls another through an API with a clear contract. When 1 module breaks, 3 others keep running. When 1 module needs changing, nothing has to be rebuilt.
3. A human reviews at high-risk points. We learned the Klarna lesson of 2024. Don't replace people 1:1. Each module has an AI confidence threshold; below it, the case goes to a human automatically. Gold transactions above 500,000,000 VND still need the shop owner's approval. AI never decides alone.
4. MONA uses it ourselves first, then sells to clients. All 3 modules (AI SDR, AI customer care, AI compliance) ran on the MONA Group's own operations before going to clients. We don't sell what we haven't used. That is a commitment we've kept since 2016.

In closing. And waiting for Part 2

A reference for those who want to go deeper into agent engineering: Anthropic's "Building effective agents" post describes multi-step agent design patterns similar to how MONA built Mona EduCenter's grading and Mona JMS's transaction handling. We often check these patterns against a client's custom spec before writing code.

The 3 cases you just read are 3 of about 30 long-standing clients at mid-market and above that MONA has fully automated. Custom-written software, modules linked in a closed loop, running for 2-5 years without breaking. That is the difference between "automating because it's trending" and "fully automating to truly pull ahead." Read more perspective in enterprise AI audit for a frame of reference. Read more perspective in automated English center management software for a frame of reference. Read more perspective in how business automation workflows differ from n8n / Zapier for a frame of reference.

You'll notice one thing: all 3 industries, education, beauty and gold, are not "tech-savvy" industries. Yet the owners who dared to invest in real automation software are 2-3 years ahead of their competitors, while 73% of Vietnamese businesses are still stuck at the pilot stage. The post when a business should stop hiring and automate is recommended further reading.

We at MONA will keep sharing in Part 2. The next 3 cases cover pharmacy chains, e-commerce warehouses, and gym and fitness chains. Each industry has its own pain, but the same MONA formula: write real software, redesign the process, and use it ourselves before selling.

👉 Follow the Business automation to read Part 2 as soon as we publish it. If you want a free audit of which parts of your business can be automated, contact us through the website mona. software or MONA's Zalo OA. We'll tell you plainly which parts can be automated and which parts you shouldn't touch yet, based on the experience of 30 mid-market long-standing clients. See details in the post a mid-market AI rollout framework with no vendor lock-in.

You deserve a business that runs lightly, without bureaucracy, and without depending on someone who "clocks out." We at MONA are committed to going there with you.