Using Wago
Workflows
Designing the conversation, every kind of step, the main workflow and testing.
A workflow is a conversation drawn as a graph: a Conversation starts step, then steps that send, ask, decide and look things up, joined by lines. Wago runs it for every incoming message.
Creating a workflow
- Open Workflows and create one with a name and a description. It starts as a draft.
- Open it in the designer: the palette of steps on the left, the canvas in the middle, the selected step's properties on the right.
- Drag steps onto the canvas, or press the + on any outcome to add the next step there. Every branch has its own line out.
- Save. Saving keeps the graph; it does not make the workflow live.
- Back in the list, choose Make this the main workflow.
The designer has undo and redo, copy and paste, drag-to-attach, and a variables panel listing every value the workflow collects. On a phone the palette and properties open as drawers.
The main workflow
One workflow in the organization is the main one, and it answers every message — there are no keywords or triggers to set up. Making another workflow main replaces it. A new or duplicated workflow is never main until someone makes it so.
Each WhatsApp number can also be given its own workflow under Settings → WhatsApp connection → Answers with; a number without one is answered by the main workflow.
Every kind of step
| Step | What it does |
|---|---|
| Conversation starts | Where the workflow begins — every message from a contact who is not already mid-conversation, when this is the main workflow. |
| Send a message | Plain WhatsApp text. Supports {variables}. Only inside the 24-hour window — outside it, use Send a template. |
| Ask a question | Send a question and wait for the reply. Offer reply buttons or a list and each choice gets its own line out. |
| Send a template | An approved WhatsApp template with its {{1}} placeholders filled in. The only thing you may send outside the 24-hour window. |
| Send media | An image, video, document, audio file or sticker, with an optional caption. |
| Send a location | A pin the contact can open in their maps app — your shop, a pickup point, a branch. |
| Send products | One product or a short list from your catalogue, as a WhatsApp product message. |
| Answer from catalogue | Work out which product the contact is asking about — by AI, from your Catalogues — and reply with its name, price, description and picture. Branches on whether one was found. Branches: found, not found |
| Save to DataMold | File what this conversation collected into one of your DataMold forms — a lead, an order, a booking — or look a row up. Needs DataMold connected under Settings → Integrations. |
| CalendarBook | Book an appointment in your CalendarBook calendars — the contact picks a day and a time from WhatsApp lists — or find or cancel their bookings. Branches: booked, found or cancelled, and a line for everything else. Needs CalendarBook connected under Settings → Integrations. |
| Condition | Branch true/false on something the conversation collected, or on the contact’s own details. Branches: true, false |
| Match | One outgoing line per value you list, plus a “no match” line for everything else. |
| Set a value | Work out a value and keep it in a variable. Maths, joining text, or just something fixed. |
| Call an API | Call your own system mid-conversation: your headers, your JSON body, and the answer pulled apart into variables. |
| AI reply | Let AI answer in your words, optionally from your knowledge base, and send that answer as the reply. |
| AI answer | Work an answer out with AI and keep it in a variable. You write the prompt; the answer is stored, not sent. |
| Search knowledge | Find the passages in your knowledge base that answer something, and keep them in a variable for an AI step below. |
| Update the contact | Save what this conversation learned onto the contact — a name, an email, a tag — so the next one starts knowing it. |
| Pause | A short beat before the next message, so two messages do not land as one wall of text. |
| Run a workflow | Run another workflow here, then carry on with this one. Build a greeting or a menu once and reuse it. |
| Hand to a person | Stop replying automatically and flag the conversation for a person. Nothing below runs until somebody picks it up. |
| End | Send an optional sign-off and close this conversation. The next message from the contact starts a workflow afresh. |
Variables
An Ask step saves the answer under a name, and any later message can use it in braces — {name}, {city}. The contact's own details are always there as {party.name}, {party.phone} and so on; on WhatsApp the number is the
identity, so there is nothing to look up first.
What a run collects is remembered: if the same customer writes again within 24 hours, the workflow starts from the top with their answers still filled in. After 24 hours of silence it starts fresh.
Let AI draft it
Describe the conversation in plain words — “greet them, ask whether they want to order or book, answer orders from the catalogue and hand bookings to a person” — and the AI composer lays out the steps for you. It knows exactly what each step can do, the same description the properties panel uses. Review and adjust what it drew, then save.
Testing before it is live
The Test chat runs the real workflow engine with a pretend customer. Type as they would and watch each reply. Nothing is sent to WhatsApp and no messages are counted, so test as much as you like. Steps that reach outside — Call an API and the AI steps — really run, so what you see is what a customer would get; AI steps use tokens from your allowance.
Built-in safety limits
Every run has a limit on how many steps it may take, how many messages it may send and how deeply one workflow may run another, so a loop drawn by mistake cannot flood a customer or your allowance.