What is human-in-the-loop

Human-in-the-loop (HITL) is a work model where artificial intelligence does not act alone: it proposes, suggests, or automates part of the process, but there is always a person who reviews and approves the final result before it is executed. In the context of WhatsApp Business, this means that AI drafts responses for your customers, but you (or someone on your team) accept them with a click before sending.

Why it matters to an SME

If you have a small or medium-sized business, you probably can't hire a large support team, but you also don't want a bot to respond with anything and damage your reputation. The human-in-the-loop approach gives you the best of both worlds: the speed and assistance of AI for drafting, and human control and judgment to avoid mistakes. This way, you can handle more inquiries in less time, without losing the personal touch.

Concrete example in WhatsApp Business

Imagine a customer writes to you at 9:00 PM asking if you ship to another province. AI instantly suggests a clear and friendly response: "Hello! Yes, we ship nationwide. The cost depends on the postal code, could you send me your address to calculate it?".

You read it, change the tone if necessary, approve it with a click, and that's it. The customer receives a response in seconds, but with your supervision. If the inquiry is more complex, AI can also help you draft a response, but the final decision is always yours.

Common mistakes when implementing HITL

• Blindly trusting AI without reviewing: although suggestions are usually good, they can fail with odd questions or sarcasm. Always review before sending.
• Not training the team: if your employees don't know how to approve or correct suggestions, the system won't work. You need to explain that AI is a help, not a replacement.
• Thinking it's only for support: human-in-the-loop also works for sales, collections, and campaigns. Use it in all areas that use WhatsApp.

How it's used in practice: a complete example

Let's consider a real case: a real estate agency in Buenos Aires receives 50 inquiries per day on WhatsApp. Without HITL, an agent would take about 5 minutes per inquiry, and by 5:00 PM they're already overwhelmed. With a platform like Wando, the AI reads each message and suggests a response based on what the business has previously answered. The agent approves it with a click and moves on to the next one.

For example, a customer asks: "What requirements are needed to rent?" The AI suggests: "Hello, to rent you need an ID, pay stubs, and a guarantor. Would you like me to send you the full list?" The agent edits it if necessary and sends it. So, instead of 50 responses from scratch, they are 50 approvals. The time per inquiry drops to less than a minute.

Additionally, the AI learns from each correction. If the agent changes "guarantor" to "security deposit insurance", the next suggestion already takes that into account. There's no need to program flows: the system adapts on its own to how the team works.

What it's confused with and how it differs

HITL is confused with two things: the automatic chatbot and flow automation. A classic chatbot responds on its own, without supervision, and if it makes a mistake, the customer is left with a bad response. A programmed flow (like "if they say hello, tell them this") works well for what's planned, but breaks as soon as the customer asks something off-script.

Human-in-the-loop is the middle ground: the AI proposes, the person decides. It's not a bot that acts on its own, nor a rigid template that needs maintenance. It's an assistant that learns from your real responses and leaves you with the final control.

Another common confusion is thinking that HITL is only for support. It also works for sales: the AI suggests responses to questions about price, availability, or shipping, and the salesperson quickly approves them. This way you can close more sales without adding staff.

What happens if you ignore it

If you don't use HITL, you have two options: handle everything manually or let a bot respond on its own. The first consumes your hours and makes customers wait. The second is a risk: an unsupervised bot can give wrong information about prices, hours, or availability, and that turns against you.

A customer who receives an incorrect response doesn't write again: they go to the competition. And on WhatsApp, word of mouth is immediate. A single bad response can cost you a sale, and several can cost you the business's reputation.

Human-in-the-loop avoids that dilemma. It gives you the speed of AI without the risk of automatic error. If you don't implement it, you'll keep losing time or customers, or both.

How it relates to Meta's official API

Meta's official WhatsApp API (WhatsApp Cloud API) is the channel through which platforms like Wando connect to WhatsApp Business. It's not the same as the WhatsApp app: it's an interface for developers that allows sending and receiving messages programmatically.

Human-in-the-loop works on top of that API. The platform receives the customer's message, the AI generates a suggestion, and when you approve it, the platform sends it through the API. Without the API, no integration is possible.

An important detail: Meta's API has rules. Message templates (HSM) can take up to 48 hours to be approved, and conversation prices vary by country. It's not something you have to handle yourself: the platform takes care of it, but it's good to know it exists so you're not surprised if a campaign is slow to start.

When NOT to use human-in-the-loop

HITL is not for everyone. If your business receives fewer than 10 messages per day, you might not need a platform: you can handle them yourself without AI help. The tool adds value when volume starts to overwhelm the team.

It's also not ideal if you want total automation, like a bot that answers frequently asked questions on its own at 3 in the morning. In that case, you need a chatbot with programmed flows, not HITL. Some platforms offer both modes, but if your priority is that no one reviews, HITL is not the answer.

And if your team is not willing to review suggestions, the system won't work. AI needs feedback to learn. If no one approves or corrects, suggestions don't improve and time is wasted anyway.

Common mistakes of those who are just getting it

The first is thinking that AI will solve everything on its own. HITL is not magic: it's a tool that saves you time, but you need to review. If you approve without reading, you're going back to an automatic bot, but with more steps.

Another mistake is not properly configuring business hours. If AI suggests responses outside your hours and no one approves them, the customer still waits. Platforms like Wando allow auto-responses by schedule, but you have to activate them.

There's also the temptation to use HITL for everything, even for messages that don't need it. A response to "hello" doesn't require AI. Use it for what adds value: repetitive questions, price inquiries, shipping, hours. That's where the real savings are.