10 challenges, 3 levels, one dataset each

Furniture showrooms, agencies and training centres miss enquiries or reply too late, so warm leads go cold.

1. Easy~20 min

   A form captures the enquiry → save it to Google Sheets → notify sales.

2. Medium~40 min

   AI reads the request, classifies the service and its priority (hot / warm / cold), then drafts a tailored first reply.

3. HardStretch

   Route each lead to the right salesperson by specialty, flag duplicates (same phone or email), and send a reminder when a lead is still untouched after the SLA.

Shops take orders through free-text chat. Staff re-type every order and often get products or quantities wrong.

1. Easy~20 min

   An order form → look up the price list → calculate the total and generate an order confirmation.

2. Medium~40 min

   AI turns a Vietnamese chat message into a structured order (items, quantity, phone, address) and asks the customer when details are missing.

3. HardStretch

   Check stock, suggest a substitute when an item is sold out, and wait for the seller's approval before recording the order.

Receivables live in Excel. Reminder dates get forgotten and cash comes in late.

1. Easy~20 min

   Read the receivables sheet → filter invoices that are due → create reminder emails.

2. Medium~40 min

   Group invoices into due soon / overdue / paid, and write a reminder in the right tone for each group.

3. HardStretch

   Match receivables against sample bank transactions, mark the matches, send unclear cases to accounting, and stop reminders for invoices already paid.

Complaints arrive scattered across channels, and nobody knows what to handle first.

1. Easy~20 min

   Receive feedback → create a ticket ID → log it and notify the responsible team.

2. Medium~40 min

   AI classifies each case as delivery, return or warranty, detects urgency and drafts a reply.

3. HardStretch

   Check the return and warranty policy to recommend a resolution. Escalate to a person when the policy doesn't cover the case, and send reminders on overdue tickets.

Shops and cafés only notice a shortage when a customer tries to buy.

1. Easy~20 min

   Read the stock sheet → alert on items below their minimum threshold.

2. Medium~40 min

   Use average daily sales and supplier lead time to estimate the stock-out date and suggest how much to reorder.

3. HardStretch

   Group the suggestions by supplier → draft purchase requests → a manager approves them before the status is updated.

Staff spend hours confirming bookings, and customers forget appointments or get double-booked.

1. Easy~20 min

   Receive a booking request → save it → send a confirmation.

2. Medium~40 min

   Check free slots, offer alternative times when there is a clash, and send a reminder before the appointment.

3. HardStretch

   Allow rescheduling and cancellations. When a slot opens up, invite people on the waitlist and confirm only the first one who accepts.

Employees send invoices through many channels. Accounting re-types them by hand and details are often missing.

1. Easy~20 min

   A payment-request form with invoice upload → store the file → notify accounting.

2. Medium~40 min

   AI extracts the supplier, date and amount from the image or PDF → writes them to the sheet and flags fields to double-check.

3. HardStretch

   Detect likely duplicates, check spending limits using sample rules, and route each request to the right approver.

SMEs have no marketing staff and spend hours writing product and promotion posts.

1. Easy~20 min

   Enter product details → AI writes a Vietnamese post → save it as a draft.

2. Medium~40 min

   Create versions for Facebook, Zalo and email in the brand voice, and check prices and offers against the input data.

3. HardStretch

   Add an approve / request-changes loop and scheduling. Only approved content moves to the publishing queue.

Shops selling electronics, computers and devices struggle to track repair requests and assign technicians.

1. Easy~20 min

   Receive a request → create a warranty ticket → send the customer a tracking code.

2. Medium~40 min

   Check the purchase date against a sample policy to decide warranty status, then classify the fault type and area.

3. HardStretch

   Assign a technician by area and skill, send reminders on late jobs, and update the customer on progress.

Owners have to open several files to understand revenue, orders, receivables and open issues.

1. Easy~20 min

   Aggregate today's sales → send revenue, order count and average order value.

2. Medium~40 min

   Compare with yesterday or the target, and let AI explain the changes in plain Vietnamese.

3. HardStretch

   Combine revenue, receivables and stock → recommend the top three priorities, each backed by numbers.