Building a WhatsApp Chatbot with Dialogwa

When people ask us for a WhatsApp chatbot, they expect a wall of code and a developer booked for a month. We disagree. The first chatbot most businesses need is a simple menu tree that answers the top five questions without a human in the loop. You do not need AI for that. You need structure, and structure is exactly what we give you. Here is how we make a useful bot possible in an afternoon.

The dialog is a tree, not a script

We model conversations as a hierarchy of nodes. Each node has text, an optional media file, and child nodes. When a customer replies with a number, we walk down the tree. This is enough for helpdesks, FAQs, catalogs, and booking flows, and it stays editable by anyone on your team. A marketer can reshape the tree without asking engineering, which is the whole point.

What the data looks like

Under the hood a dialog is plain JSON. You can build it in our UI or push it through the API, which means version control and code review are options, not dreams:

{
  "dialog": "Welcome. Reply with the number you need.",
  "child": [
    { "dialog": "What is Dialogwa?", "title": "About",
      "child": [{ "dialog": "We are a WhatsApp gateway you control." }] },
    { "dialog": "Pricing", "title": "Plans",
      "child": [{ "dialog": "Plans start free for 7 days." }] }
  ]
}

Building it in the UI

1. Open the Dialog menu and pick your session.
2. Write the root welcome message. Keep it to one screen.
3. Add child nodes for each menu option, then nest deeper as needed.
4. Toggle the dialog active and watch replies flow.

When to reach for code

The tree covers most cases, but sometimes you need live data: order status, account balance, stock levels. That is where a webhook beats a static node. We post the inbound message to your URL and you return the text. You keep the friendly menu for everything else, and your database stays behind your firewall where it belongs.

What we learned shipping this

Our earliest users tried to model their entire company in one tree. It failed. The trees that worked had five or fewer top-level options and a human escape on every branch. We now steer new builders toward that shape because it survives contact with real customers.

The caveat we always give

A tree with twenty branches and no exit is worse than no bot. Customers get lost and blame your brand, not the menu. We recommend short top-level menus, a clear way to talk to a human, and a yearly review of which branches nobody clicks. Build for the question people actually ask, not the org chart.

Test with real customers, not yourself

The trap we see most is founders testing the bot by sending messages they already know the answer to. That confirms the tree works; it does not confirm the tree helps. We ask new builders to hand the number to three real customers and watch without intervening. Where they hesitate, where they type words the menu never offered, those are the gaps to close. A bot earns trust in the first confusing moment, not the happy path, so test the confusion first.

Keep the first message honest

We tell every new builder the same thing about that welcome menu: do not oversell. A bot that promises instant answers and then stalls on question two teaches the customer that automation here is theater. We would rather the first message say "we answer the common stuff fast, and a person takes the rest" because that is true, and true scales. The brands that win on WhatsApp are the ones whose bot and human feel like the same team, not rivals.

Our suggestion

Ship the smallest tree that removes your most repeated question. Measure how often it is used, then expand only the branches that earn their place. A bot is a garden, not a monument.