A WhatsApp chatbot in Dubai usually means one of two things: a menu-based bot that follows fixed rules, or an AI agent that reads what the customer actually typed and answers in context. The difference shows up the moment a customer writes a full sentence instead of picking option 2. Two sample transcripts below show which suits your business. For the wider picture, see our guide to AI WhatsApp agents for UAE businesses.
How menu-based chatbots work
A menu-based chatbot is a decision tree. The business writes every path in advance, and the customer moves through it by sending a number or a keyword. Many chatbot tools also show numbered menus or buttons to make the choices easier to pick.
A typical flow looks like this:
- The customer says hello.
- The bot replies with a list: 1 for prices, 2 for opening hours, 3 for location, 4 to speak to staff.
- Each number leads to a fixed answer or a further menu.
This works well when questions are predictable and few. It is easy to test, because every reply is written by a person. The weakness is everything outside the tree. If a customer writes “do you deliver to JLT tomorrow before 10”, a rule-based bot either matches one keyword and gives a partial answer, or falls back to “Sorry, I did not understand”.
How AI agents handle free text
An AI agent reads the whole message, works out what the customer wants and replies from the information the business has given it: products, policies, opening hours and past questions. It does not need the customer to follow a menu.
In practice, a well-set-up agent can:
- understand a question written in English, Arabic or a mix of both;
- handle two or three questions in one message;
- remember what was said earlier in the same chat, so “and the bigger size?” makes sense;
- ask a short follow-up question when something is unclear;
- hand the chat to a person when the question is outside its knowledge or needs a human decision.
The quality of the answers depends on the quality of the knowledge you give it. An AI agent with thin or out-of-date information will still give thin or out-of-date answers, just more fluently.
Same question, two transcripts
Here is the same enquiry to an imaginary bakery, first with a menu bot and then with an AI agent.
Menu-based chatbot
Customer: Hi, do you have sourdough without sesame? Need 2 for Friday morning, delivery to Al Barsha
Bot: Welcome. Please choose: 1 Menu, 2 Opening hours, 3 Location, 4 Talk to us
Customer: 1
Bot: Here is our menu: [link]
Customer: but is there one without sesame?
Bot: Sorry, I did not understand. Please choose 1, 2, 3 or 4.
AI agent
Customer: Hi, do you have sourdough without sesame? Need 2 for Friday morning, delivery to Al Barsha
Agent: Hello, yes. Our plain sourdough and rye sourdough have no sesame topping. Both are baked in a kitchen that also handles sesame, so please let us know if this is for an allergy. Friday morning delivery to Al Barsha is available. Would you like 2 plain, 2 rye or one of each?
Customer: one of each
Agent: Noted: 1 plain and 1 rye for Friday morning, Al Barsha. Shall I send you the basket with the payment link?
The menu bot was not wrong, it simply could not cope with a normal sentence. The AI agent answered all three parts, flagged an allergy question honestly and moved the customer towards an order.
Where rules still matter
AI does not remove the need for rules. It changes where you put them. Good agents work inside clear limits set by the business:
- Facts it must not change: prices, delivery areas, opening hours and policies come from your information, not from the model’s guesses.
- Topics it must not handle alone: complaints, refunds, medical or legal questions and anything involving a judgement call go to staff.
- Tone and wording: greetings, sign-offs and phrases to avoid, written down in a brand voice sheet.
- Fixed messages: some messages should be identical every time, such as payment instructions or a returns policy summary.
These rules are written in plain language, for example “never promise same-day delivery after 4pm” or “if a customer mentions an allergy, recommend they confirm with staff”. You can see how settings like this are organised in our help article on configuring the AI assistant.
Choosing for your business
Use this table as a quick check.
| Your situation | Menu bot | AI agent |
|---|---|---|
| Fewer than five common questions, always phrased the same way | Often enough | May be more than you need |
| Customers write long messages or mix questions | Struggles | Suits this |
| Customers write in Arabic, English or both | Needs a separate tree per language | Replies in the customer’s language |
| Large catalogue or many services | Menus become long | Answers from the catalogue |
| You want orders or appointment requests from the chat | Possible with long flows | Handled in conversation |
| Strictly scripted answers required | Suits this | Needs fixed-message rules |
A simple test: read your last 50 WhatsApp enquiries. If most of them could be answered by picking from a short list, a menu may be enough. If most are full sentences with details, an AI agent will save your team more time.
How Dub Ai helps
Dub Ai puts an AI sales agent on your WhatsApp number that answers customers in their language, recommends products and builds baskets or takes appointment requests your team confirms. It answers from your own knowledge: products, prices you have set in your store, policies and FAQs. You decide the rules, the tone and the topics that go straight to staff, and your team can take over any chat from the shared inbox. Dub Ai is a business-specific assistant for your own customers, orders and enquiries, not a general-purpose chatbot. Our team sets it up with you and adjusts it as you review real conversations.
Frequently asked questions
Is an AI agent harder to control than a chatbot?
It is controlled differently. A menu bot is controlled line by line, while an AI agent is controlled through its knowledge, written rules and handover triggers. With clear rules and a weekly review of real chats, most businesses find it predictable. The key is to keep sensitive topics, such as refunds or complaints, with staff rather than asking the AI to decide.
Can I combine buttons and AI replies?
Many chatbot platforms let you mix fixed menus with free-text AI answers, for example a short menu at the start and AI for everything typed. Whether that suits you depends on your customers. If they mostly type full questions, a menu adds a step. If you need a fixed choice, a numbered list in the message works on any setup.