What is a prompt
A prompt is an instruction or set of instructions given to an artificial intelligence model to perform a specific task. It can be a question, a command, a context text, or a combination of both. In the WhatsApp Business ecosystem and tools like Wando, prompts are used to configure how the AI suggests responses to your customers' messages.
Why it matters to you as an SME
If you have a business and use WhatsApp to serve customers, a good prompt can make the AI propose useful responses with your tone, saving you time. A poorly written prompt, on the other hand, generates generic or outright wrong suggestions. It's not magic: the AI reads your prompt and uses it as a guide to respond. That's why the clearer and more specific you are, the better results you'll get.
Concrete example
Imagine you have a clothing store and a customer asks: 'Do you ship nationwide?' A basic prompt would be: 'Respond to customers in a friendly manner.' A better prompt would be: 'You are the virtual assistant of a clothing store. Respond in a cordial tone, in Spanish, and if they ask about shipping, tell them we ship nationwide with a cost of $1500. If they ask about something you don't know, offer to transfer to a human.'
The difference is huge: the second prompt gives the AI context, rules, and limits. In Wando, these prompts are used to train the AI to learn from your real responses and improve over time, always under your team's supervision.
Common mistakes when writing prompts
- Being too vague: 'respond well' is not useful. You need to specify the tone, information, and limits.
- Providing incorrect information: if the prompt says shipping is free and it isn't, the AI will make mistakes.
- Not updating the prompt: if you change prices or hours, the prompt must be updated so the AI doesn't repeat old data.
- Omitting the 'I don't know': without a clear limit, the AI can invent responses. Always tell it what to do when it doesn't know something.
Remember: AI suggests, but the final decision is always made by your team. At Wando, each suggested response is approved with one click before it is sent.
How it works in practice: a complete example
Imagine a restaurant in Buenos Aires. It receives via WhatsApp: 'Do you have a table for two tonight?'.
With Wando, the owner sets up an initial prompt: 'You are the reservations manager of a restaurant. Respond in a friendly tone, confirm availability if possible, and if not, offer alternative times. If they ask about the menu, mention that we send it via WhatsApp.'
When the message arrives, the AI suggests a response: 'Hello! Yes, we have a table for two at 21:00. Shall I confirm it for you?' The owner approves it with one click and it is sent. If the customer asks something the AI doesn't know, the owner writes the correct response and the AI learns from that exchange for future queries.
This flow is human-in-the-loop: the AI proposes, the team decides. This prevents a bot from answering anything and builds a history of real responses that improves suggestions over time.
What it is confused with and how it differs
People often confuse a prompt with a flow chatbot. A traditional chatbot is programmed with branches: if the customer says A, respond with B; if they say C, respond with D. That requires constant maintenance and breaks when the customer asks something unexpected.
A prompt, on the other hand, is a guide in natural language. You don't define every path: you give it context, rules, and limits, and the AI interprets. In Wando, moreover, the prompt is not static: the AI learns from the real responses that your team approves, so it adapts without you having to reprogram anything.
Another common confusion is thinking that a prompt is only for chatbots. It is also used to generate campaign texts, summarize conversations, or prepare answers to frequently asked questions. The key difference is that the prompt does not replace human judgment: it suggests, it does not decide.
What happens if you ignore it: the concrete consequence
If you don't use prompts, the AI in your WhatsApp inbox works blindly. Without context, the suggestions are generic: 'Thanks for writing to us!' for any query. That doesn't help anyone.
The concrete result is that your team wastes time editing each response, and customers feel they are talking to a robot. In the worst case, the AI makes up information: an old price, a schedule that no longer exists, a canceled promotion. That generates complaints and damages trust.
With a minimal prompt, on the other hand, the AI already knows what to say and what not to say. It's not magic: it's giving the tool the context it needs to be useful. Ignoring it is leaving money and reputation on the table.
How it relates to Meta's official API
Wando uses Meta's official WhatsApp API (WhatsApp Cloud API). That means messages travel through Meta's infrastructure, with approved templates for campaigns and messages outside the 24-hour window.
Prompts do not replace the API: they use it. When the AI suggests a response, that response is sent through the Meta API. Templates (HSMs) used for campaigns can also be configured with prompts, so that the AI generates the initial text.
Template approval can take up to 48 hours. Meta's conversation prices vary by country and are governed by its official documentation. Wando gives you the AI and management layer, but the connection with WhatsApp is always through the official channel.
When it does NOT apply or is not convenient
A prompt is not the solution for everything. If your business handles very technical or sensitive queries (for example, legal or medical topics), the AI may not be prepared to respond accurately. In that case, it is better to route directly to a human, without automatic suggestions.
It is also not convenient if your message volume is extremely high and you need 100% automatic responses without supervision. Wando is designed with human supervision: the AI suggests, the team approves. If you expect it to respond on its own all the time, there are other tools, but with risk of errors.
And if you don't have time to write a good initial prompt, better not to start: a poorly made prompt generates more work than it saves. The good news is that in Wando you can adjust it and the AI improves with your real responses, so the initial effort pays for itself.
Most common mistakes of those who just understand it
The first is thinking that a prompt is 'write something and done'. A good prompt requires thinking it through: role, tone, information, limits. The second is not updating it: if you change prices or hours, the prompt must reflect that, otherwise the AI repeats old data.
Another mistake is not including an 'I don't know'. Without a clear limit, the AI can invent answers. Always tell it what to do when it doesn't know something, for example: 'If you don't know, offer to transfer to a human'.
There's also confusion between the prompt and programming: you don't need to know code, but you do need to think about common scenarios. And finally, believing that AI replaces judgment: at Wando, the final decision is up to the team, and that's an advantage, not a limitation.
Frequently asked questions
Is a prompt the same as a command?+
In general, yes: a prompt is an instruction you give an AI to perform a task. It can be a question, an order, or a text with context. In tools like Wando, prompts are used to configure AI behavior in the WhatsApp inbox.
How do I write a good prompt for my business?+
Start by defining the AI's role (for example, 'you are the assistant of a clothing store'), then the tone (friendly, formal, etc.), then key information (prices, hours, policies), and finally rules for when it doesn't know something. Test and adjust based on results.
Are prompts only used for chatbots?+
No. They are also used to generate texts, summarize conversations, create responses for campaigns, and more. In the WhatsApp Business context, prompts help the AI suggest useful responses that your team then approves and sends.
What happens if I write a bad prompt?+
The AI can give incorrect or unhelpful responses. That's why it's important to review and update prompts periodically, especially when your products, prices, or policies change. In Wando, the AI learns from your real responses, so the better the prompt, the better the learning.
Can I use the same prompt for all my channels?+
In Wando, you can configure prompts for WhatsApp, Instagram, and Messenger. Although the tone can be similar, each channel has its own context. For example, on Instagram people are more informal. We recommend adjusting the prompt per channel so suggestions are more natural.
What do I do if the AI suggests an incorrect response?+
You write the correct response yourself and send it. Wando learns from that correction and uses it for future suggestions. It's part of the design: the AI improves with your real responses, not with theory.
Are prompts stored in a visible place?+
Yes, in Wando's settings you can view and edit your prompts at any time. You don't need to know how to code: it's plain text that you can adjust as your products or policies change.