Escalation to a human agent

Human escalation is the process of transferring a conversation that was being handled by a bot or an artificial intelligence to a human agent, when the AI cannot resolve it or the customer explicitly asks for it. In the WhatsApp Business ecosystem, this is key to not losing sales or customers: the machine handles the simple stuff and routes the complex to a person.

Why it matters to your SME

If you use a chatbot or an AI assistant, human escalation is the moment of truth. A well-configured AI resolves frequent queries (hours, prices, order status) and escalates to a human when it detects anger, complex doubts, or a sales opportunity that requires personal attention. Doing it well improves the customer experience and prevents them from hating you for talking to a robot.

Concrete example

Imagine an online clothing store. A customer writes: "Do you have the black t-shirt in size M?". The AI responds right away: "Yes, we have it, $15,000". But then they write: "And if it doesn't fit, can I exchange it? It happened last time that they charged me for the return shipping". At that point, the AI should escalate to a human, because there is a previous complaint and a situation that requires judgment. If it doesn't escalate, the customer gets frustrated and you might lose the sale.

Common mistakes

• Escalating everything: if you route every query, the AI is useless and your team gets overwhelmed.
• Never escalating: letting the AI improvise on sensitive topics (complaints, billing, legal issues) is dangerous.
• Escalating without context: passing the chat to a human without a summary of what happened forces the customer to repeat everything.

How it works in practice, with a complete example

Let's consider a real case of an aesthetics clinic in Córdoba. The owner configures Wando with Meta's official API and loads her services and promotions into the inbox. A client writes at 9 PM: "Do you have an appointment for tomorrow for waxing?".

The AI suggests a response with the available times and the client confirms. Then she asks: "And if it rains, can it be rescheduled? Will I be charged anything?". At that point, the AI has no criteria to respond about the cancellation policy: the team approves the suggestion or takes direct control. The chat escalates to a human with all context visible.

The important thing is that the escalation does not cut off the conversation: the human sees the history, responds, and the AI continues learning from that response for the next time.

What it is confused with and how it differs

Human escalation is confused with two things: with the traditional chatbot and with the simple manual chat handoff.

A traditional chatbot (like ManyChat) works with programmed flows: if the client writes a keyword, it goes down one path; if not, it falls through. Escalation there is just another rule: if no one responds, it forwards. But there is no learning or intelligent context.

Wando is not that: the AI suggests responses based on the ones you have already approved. And escalation is not a leap into the void: it is passing control to a human with a summary of the conversation and response suggestions. The difference is that the AI helps you escalate well, not just escalate.

What happens if you ignore it

If you don't escalate to a human when needed, the cost is not just an angry customer: it's real money. A sales inquiry that the AI can't resolve is lost; a complaint that is handled poorly turns into a refund, chargeback, or public comment.

In SMBs in LatAm, where WhatsApp is the main sales channel, a customer who gets frustrated with a bot doesn't come back. And on top of that, they tell others.

Ignoring escalation also leaves you relying on the AI always being right, which doesn't happen. Without a clear handoff mechanism, your team ends up handling everything in a rush, or the bot improvises on sensitive topics like pricing or warranties. Neither is good.

How it relates to Meta's official API

Human escalation in WhatsApp Business relies on Meta's official API (WhatsApp Cloud API). That API allows you to receive and send messages, but it doesn't decide for you: it has no business logic or context.

Tools like Wando connect to that API and add the intelligent layer: the AI that suggests responses, the rules for when to escalate, and the dashboard where the human takes control. Without the API, you couldn't escalate with context or comply with Meta's policies.

One detail: Meta's conversation prices vary by country and are updated. If your business handles volume, it's worth checking Meta's official documentation to avoid surprises. The API is the foundation; escalation is the strategy.

When it does NOT apply or is not convenient

Human escalation is not for everyone. If your business receives few messages and they are handled by a single person, you might not need an AI tool: the WhatsApp Business app is enough.

It is also not convenient if your operation is 100% transactional and does not require judgment: for example, a bot that only confirms orders and does not handle complaints. In that case, escalation adds friction.

And be careful: if your team is not willing to supervise AI suggestions, the tool loses its purpose. Human escalation works when there is someone on the other side willing to take control when needed. Otherwise, it is just a chatbot with a different facade.

Common mistakes of those who are new to this

The biggest mistake is thinking that escalating to a human is the same as turning off AI. No: it is passing control to a person with the context and tools to resolve.

Another classic: setting overly broad escalation rules (e.g., "any message with the word complaint") and ending up escalating everything. That overwhelms the team and cancels out the benefits of AI.

And the third: escalating without context. If the human receives the chat without knowing what happened before, the customer repeats everything and the experience dies. The solution is to use tools that pass on the history and suggestions, like Wando does with its inbox.