Chatbots are automated programs that chat with customers via text — answering questions, providing information, and performing simple tasks based on scripts or AI. Customers text, chatbots respond immediately 24/7 on Zalo, Facebook, website without the need for on-duty staff. Chatbots are divided into two large groups: script-based chatbots (rule-based, runs according to menus and installed keywords) and AI chatbots (understands natural language, responds flexibly according to trained content). Zenify records AI chatbots handling an average of 74% of level 1 requests on customer systems.
TL;DR
- Chatbot = automatic chat program, instant response 24/7.
- 2 main groups: rule-based (fixed scenario) and AI (free question understanding).
- Also divided by location: website, Facebook, Zalo, social networks, internal.
- Choose AI chatbot when customers ask diverse questions; Select the scenario when the process is fixed.
- Chatbot automatically handles ~74% of level 1 requests on the Zenify system (data needs to be confirmed).
How does Chatbot work?
Chatbots operate according to two main mechanisms — scripted or AI — and you need to understand them correctly to choose the right one.
Rule-based (script): runs according to the preset question branch. Customers click the menu or type keywords; bot follows the corresponding flow. Advantages: easy to predict, cheap, not "made up". Disadvantages: deviating from the script means giving up.
AI chatbot: understands natural language thanks to language models, answers questions freely based on the knowledge base. Advantages: flexible, less time writing scripts. Disadvantage: needs supervision to avoid incorrect answers.
The most popular mechanism today is a combination: AI understands the question, the script takes care of the standard transaction (getting information, transferring people), and there is always a button to hand over staff when needed.
5 popular types of chatbots
There are 5 types of chatbots you will encounter when choosing a tool:
Type | Features | Use when
Chatbot script (rule-based) | Runs by menu/keyword, easy to predict | Fixed process: ask price, schedule, FAQ
AI Chatbot (LLM) | Understand sentences freely, answer flexibly | Customers ask a variety of questions, need natural advice
Chatbot website (Livechat) | Mount on the web, generate leads from visitors | Collect leads and provide support as soon as customers read the website
Facebook Chatbot / Zalo OA | Runs in fanpage/OA, automatic messaging | Sales & Customer Service on social networks
Internal Chatbot | Support staff lookup and self-service | Internal helpdesk, questions and answers about company processes
Which one to choose depends on where the visitor is and the variety of questions — not "more expensive chatbots are better".
Are AI chatbots better than scripted chatbots?
Not really — each type is optimal for a different situation. Straight up comparison:
Criteria | Chatbot script | Chatbot AI
Implementation costs | Low, fast | Higher, need data
Answer questions freely | Poor | Good
Risk of wrong answer | Low (with script) | Must monitor (AI QA)
Maintenance | Edit each branch | Update knowledge base
Guest Experience | More rigid | More natural
The best practice for small and medium businesses is AI chatbot + combined scenario: AI handles common questions, the script takes care of standard processes (send quotes, collect information), handover button is always available. Zenify notes that this combined AI chatbot handles an average of 74% of level 1 requests on its own, with the rest handed over to staff with full context.
How to choose the right chatbot for your business?
The way to choose is to determine the goal and where customers contact first, then compare technology.
Define the goal: answer repetitive questions (→ scenario/AI consultation) or automated sales (→ scenario of gathering information + transferring people)?
Identify channel: Are customers texting on Facebook, Zalo, or visiting the website? Choose a chatbot to install in the right place.
Measuring volume: if there are only 10 news per day, a simple script is enough; If there are hundreds of messages asking all kinds of questions, you need AI.
Smooth "handover" requirement: the bot must transfer the conversation (with context) to the employee when the customer needs it — this is a required feature.
Quality monitoring: Requires reporting the rate of bots answering correctly and scoring conversations, avoiding bots silently answering incorrectly.
If you want to deploy methodically, read more: What is AI Agent? How is it different from traditional chatbots and How to build an effective Zalo chatbot for sales.
Mistakes when making chatbots
Expect bots to do it all — bots cannot replace complex consulting; Let the bot handle level 1, hand over the rest.
Not updating the knowledge base — AI bot answers incorrectly because of old data; Periodic updates are required.
No "call someone" button — customers stuck in a bot loop will be very upset; always provide a way out.
Not measured — without knowing the rate of bots processing themselves, the rate of correct answers cannot be improved.