How not to lose quality when using a chatbot to automate responses

Level: Intermediate

We had the pleasure of interviewing Jorge García, our Product Director, for several articles on artificial intelligence. During one of those informal meetings where we were discussing potential use cases in the contact centre sector, he told us:

“The philosophy behind self-service should be nothing more than automating customer service without losing the quality of experience you would get when talking to an agent.”

It’s true. The word “self-service” can have a certain negative connotation because it evokes a scenario where the customer is alone in the face of danger.

For example, we call the checkout machines in supermarkets self-service, but the machine does not help the consumer to scan the products, the consumer scans them alone.

So when we talk about an AI-powered chatbot that serves users who come to the website and provides answersto their questions or redirects them to the best available agent, we are talking about a system that automates responses.

We have been stressing this a lot lately, but we think it is very important to remind those companies that are thinking of investing in this technology that if it is not applied with certain criteria, the bot can end up truncating the customer experience.

WHAT ARE THE KEYS TO MAINTAINING QUALITY WHEN AUTOMATING RESPONSES?

1. PROCESS SIMPLE AND REPETITIVE ENQUIRIES

The first thing we can start automating are FAQs. Frequently Asked Questions are those that come up all the time, either before buying a product/service or after.

One of the questions a telecommunications company could automate would be “Why has my phone bill gone up this month?”.

If we connect the AI to our customer service agent, the bot can access the customer’s bill, analyse the document and inform the customer of any additional charges.

2. TAKING THE CUSTOMER’S EMOTIONAL STATE INTO ACCOUNT

The AI that powers the chatbot not only understands the user’s request, but also has the ability to extract their emotions. As a result, it knows from the outset the severity of the incident and the emotional state of the customer.

The bot is not a human, so no matter how much we train it to deal with serious problems, it will always lack emotional intelligence.

So if the bot is talking to a customer who has a power cut at home, for example, we need to train the AI to transfer the conversation to an agent.

3. TRANSFER THE CONVERSATION TO THE AGENT WHEN THE USER REQUESTS IT

There is nothing more frustrating for a user than talking to a bot when they want to talk to an agent. So it’s always a good idea to give the user instructions on how to get in touch with an agent.

In addition, it will soon be illegal to offer customer service based on voicemail or bots that do not allow you to speak to an operator. When the new Spanish customer service law comes into force, companies that do not comply will be fined up to €100,000.

4. AVOID BOT HALLUCINATIONS AT ALL COSTS

We don’t want our bot to give a customer the wrong answer. So we need to learn how to maintain the knowledge bases properly.

When we create a knowledge article, we must avoid explaining multiple processes in the same document.

WANT TO ADD A CHATBOT TO YOUR BUSINESS?