What you actually pay for
You pay for work, not for the AI model itself. The model is rented by usage, and its cost is usually the smaller part of the bill.
Most of the build effort goes into four things. The chatbot must know your answers, reach your systems, stop at the right moment and be tested on real questions.
- Knowledge: prices, services, opening hours, policies and the questions your staff answer every day
- Channels: Zalo OA, Messenger, a website widget, and each one has its own rules and setup
- Integrations: calendar, CRM, booking or sales software the chatbot reads from or writes to
- Hand-off: rules for when a person takes over, and how the conversation reaches them
- Testing: real past questions run through the chatbot before it talks to customers
What makes the price go up or down
Scope moves the price more than anything else. A chatbot that answers questions and collects a phone number is far simpler than one that books appointments in your software.
The table shows the usual drivers and how they change the work.
Cost drivers of an AI chatbot| Driver | Lower cost | Higher cost |
|---|
| Task | Answers questions and collects contact details | Books, quotes or updates records in your systems |
|---|
| Channels | One channel, for example Zalo OA | Several channels with one shared history |
|---|
| Languages | One language | English and Vietnamese with different answers per market |
|---|
| Integrations | Google Sheets or email | CRM, booking or industry software through its API |
|---|
| Hand-off | Notify one person | Route by topic, branch or language, with working hours |
|---|
| Data | Public information only | Personal or medical data with access rules |
|---|
Our from-price covers the lower-cost column for Zalo OA, Messenger and a website in English and Vietnamese. Anything in the right column is scoped and quoted after the audit.
Costs after launch
A chatbot keeps costing money after it goes live. Plan for these lines before you compare quotes.
AI model usage is billed by the provider per amount of text processed. It grows with the number of conversations and the length of your knowledge base, so we estimate it from your real message volume on the audit.
Platform messages can have their own fees. Zalo OA plans, ZNS template messages and WhatsApp template messages follow each platform's current price list, which we check for your account on the audit.
Someone has to update the answers. Prices, staff and services change, and an outdated chatbot gives wrong answers with confidence.
- AI model usage, paid to the provider by volume
- Platform fees where they apply, for example Zalo OA plans or template messages
- Hosting and monitoring of the chatbot
- Monthly review of conversations and updates to the answers
Subscription chatbot or custom build
A subscription chatbot costs less to start and is the right choice for many small shops. It works well when you need fixed answers and a lead form in one channel.
A custom assistant costs more upfront and makes sense when the chatbot must work with your own systems. Typical examples are booking into clinic software, quoting from a price list, or writing leads into your CRM with the right fields.
Compare the total over a year, not the first invoice. Include subscription fees per channel or per seat, model usage, and the staff time spent fixing what the tool cannot do.
How to compare two quotes
Two quotes for an AI chatbot often describe different products. Ask each vendor the same questions so the prices become comparable.
- Which channels are included, and is each channel set up and tested?
- Which systems does the chatbot read and write, and through which API?
- When does it hand a customer to a person, and how does the person see the history?
- Who pays for model usage and platform messages, and how is it estimated?
- Who owns the accounts, the data and the conversation logs?
- What does monthly support include after launch?
When a chatbot is not worth the money
Some businesses do not need an AI chatbot yet. If you get a handful of messages a day and staff answer them quickly, the build will not pay back.
A chatbot also cannot fix missing information. If your prices or schedule live only in someone's head, write them down first; the chatbot needs a source to answer from.
If you are unsure, an audit is cheaper than a wrong build. It tells you whether a chatbot pays back on your volume before you commit.