Menu chatbot vs. AI customer service: what changes

Quick answer: customers aren't averse to talking to a machine; they're averse to option menus that don't solve anything. An AI agent that talks naturally, understands voice notes and free text, and actually solves the problem leaves customers satisfied, as long as the company is transparent that it's a digital assistant.

I'll be honest: I'm the guy who gets irritated when I call a company and hear "press one, press two". I stop right there. And I think a lot of you reading this do too.

That's where the belief that "people don't want to talk to a bot" comes from. But when you look closely, the rejection isn't of the bot. It's of the menu.

How a decision-tree chatbot works

Think of the classic example: you just want a copy of a bill. To get it, you press 1 for billing, then 2 for "I'm a customer", then 3 for "I want a copy", and you still have to choose whether you want it by WhatsApp or by email.

A chatbot works well when the task is choosing from fixed options. But:

if the person didn't type the right number, the conversation gets stuck;
if they sent a voice note, it ignores it and asks the same thing again;
if the question isn't in the tree, there's no way out.

That model worked from the 90s until quite recently. AI came precisely to break the tie to "option one, option two".

How a well-configured AI agent works

An AI agent has a conversation. It understands what the person wrote in their own way, understands voice notes, asks for the missing information and solves the problem. If I'm talking to a bot, but it serves me close to how a human would and solves my problem, I'm all set.

Where AI in customer service makes the biggest difference

Outside business hours

In my operation, a good share of leads come from videos. And when do people watch videos? Outside business hours. Some people message me in the middle of the night.

With AI, that person gets an answer right away, clears up their question and leaves with the next step lined up. Without AI, they wait until the next day and, in the meantime, keep looking at other companies.

Messages that aren't "business"

You know those automatic Happy Easter and Happy Mother's Day messages? I've seen the digital assistant reply with a thank you, wishing the same to the person's family, and I thought: that's a better reply than I'd have written. Well configured, AI keeps things cordial even in the conversations a human would let slide.

People who struggle with reading and writing

This is something few people consider. I've seen cases in healthcare where a large part of the audience had trouble reading and writing. The person sends a voice note, the AI replies with audio. A conversation that would take a human a long time (listen, record, listen, record) becomes fast.

Transparency: make it clear it's a digital assistant

For ethical reasons, we make it clear that it's a digital assistant. And even so, some people thank it at the end, praise the service and even send kisses. Transparency doesn't hurt the experience; what hurts it is bad service.

How to prove to a client that it works

When I present this in a meeting, at some point I say: "Look, I've shown you a lot of stuff here, but PowerPoint will take anything. In PowerPoint I can say I'm taking you to Disney."

So I let the person talk to the agent live, during the meeting. And I suggest: "Ask it to call you." It calls. The person's own perception is worth much more than any explanation of mine, because I'm biased.

Map the real questions. Which questions come in most often? They become the agent's knowledge base.
Define the goal of the conversation. What does the agent need to validate or collect before passing it on?
Keep it focused. Instruct the AI not to make small talk: if the customer drifts off, it brings them back to the topic.
Define when to hand off to a human. If the AI tried to bring the customer back and couldn't, it transfers.
Separate AI from automation. Anything that's a fixed rule (a calculation, sending a link) can be simple automation, with no AI cost.
Don't run a menu chatbot and AI at the same time. An automatic message in the middle breaks the flow and also becomes context for the AI. Keep a simple chatbot ready and switched off as a plan B: if something goes wrong, turn it on; once it's fixed, turn it off.
Be transparent. Let people know it's a digital assistant.
Test it with the decision-maker. Before showing slides, have the person talk to the agent.

Frequently asked questions

Do people really accept being served by AI?

In my experience, yes, when the AI solves their problem. What people reject is the option menu that doesn't understand what they want.

Do I need to hide that it's an AI?

No. We make it clear that it's a digital assistant, for ethical reasons, and that doesn't stop the experience from being good.

Is a menu chatbot still useful for anything?

It's useful for simple choices and as a contingency plan. Just don't leave it running alongside the AI, because one messes up the other.

Can AI handle customer service through voice messages?

Yes. It understands the customer's voice note and can reply with audio, which helps a lot for people who struggle with reading and writing.

When should the AI hand the conversation off to a human?

When the customer goes off topic and doesn't come back, or when the case requires a human decision. That rule needs to be written into the agent's configuration.

Keep reading

Want to see it in action?

The best way to understand it is to talk to a real AI customer service assistant. Talk to Aurora, the Resultados Exponenciais assistant, right now and get your questions about sales efficiency answered:

Talk to Aurora on WhatsApp

Customers don't reject AI. They reject "press 1 for billing"

Menu chatbot vs. conversational AI: the difference