Human Or Chatbot: Can You Tell The Difference?

We’ve all suffered the frustration of spending hours on the phone with customer services, switching from representative to representative to resolve a problem.

In recent years, companies have been using online self-services and chatbots to handle queries. And most users actually prefer dealing with a computer than waiting for a person, especially for routine stuff.

But sometimes the chatbot can’t cope with a complex query. A startup in Israel offers the best of both worlds with a blend of chatbot and human responses. And here’s the thing: You won’t know the difference.

If the bot doesn’t understand a query, it will seamlessly connect with an actual representative – who will be able to see what the problem is from the chatbot’s account, so the consumer doesn’t have to explain it for the 15th time. 

But when it’s back to straightforward stuff, the human can pass you back to the chatbot, like a game of ping pong, without you even realizing. You’re bouncing between a computer for the easy bits and a real person for the hard bits.

“We are helping brands build an autonomous interface that can handle 80 per cent of the repetitive tasks, while keeping a human in the loop that will always be there to seamlessly help you,” says Yaniv Hakim, Founder and CEO of CommBox. 

It builds an automated ‘master script’ for companies to their chatbots can have ‘human conversations’ with customers, in a way that makes them feel like they’re messaging an actual representative.

“Say we are working with a healthcare company, and 70 per cent of the conversations consumers have with the company are them wanting to schedule an appointment,” says Hakim. “Only 10 per cent of the conversation is about vaccines, let’s say, or actual medical issues.”

The chatbot is trained to deal with the routine parts of the conversation, pass the patient back to a human for the medical discussion, then return them to the chatbot to check details and make arrangements.

“A company typically has 20 to 50 really important use cases (subjects) that the customer converses about, so we look at those and try to determine what would give them the best experience,” says Hakim.

“For example, some topics are best conducted through WhatsApp, or a video call with a representative. After we understand all of this, we start building the script.”

Client companies digitize every form of written interaction they have with their consumers across all platforms, from their website to WhatsApp, YouTube, Twitter, and other social media platforms.  

CommBox then goes through a ‘discovery process’ with the company to decide which types of conversations should be automated (usually the most repetitive queries), and what the bot should say in each one. Conversations that are – for the time being – too much for the chatbot will be passed on to a human.

The chances are you’ve already interacted with its automated scripts, but the chatbot seemed so natural in its communication that it would be hard to tell it apart from a human. 

“In the end, companies won’t need their websites, or their IVRs (Interactive Voice Response, or computer-operated telephone system) – because they have one script that can do everything, and can redirect the consumer to an agent if it can’t solve their query,” he says. “So why bother sending people to the website to understand the problem by themselves?”

There are many advantages for the client company. It’s a cost-effective alternative to building an entire portal on the company website.

Hakim doesn’t think automated scripts will be replacing people’s jobs any time soon, but many people, especially Gen Zs, are naturally moving away from that line of work.

The average customer service representative between the ages 20 to 34 stays on the job for just over a year, according to the United States Bureau of Labor Statistics.

And the average employee turnover rate at call centers is as high as 45 per cent – at least twice as high as other departments.

“We can simply build a one-time conversational script, and then activate it on all channels, WhatsApp, email, voice, etc,” says Hakim.

Building the conversational interface for the chatbot is very easy and fast, says Hakim. It can take only a few hours in the most straightforward case, and depends primarily on how long it takes to work out what the client needs.

CommBox outperforms rival services, says Hakim, because of the seamless switch from chatbot to human. 

The AI-script supplies customer-service reps with all the information it gathered from the customer during its conversation, and offer them relevant lines to say.

When the rep feels that their services are no longer needed they can seamlessly pass the conversation back to the bot.

Hakim also says the chatbot is a quick learner. Unlike other AI it understands what it needs to do from a small number of examples.

“When you send an email, the language you write in is different from the way you text your friend, or write a post on Facebook. So our AI also acts upon ‘non-bot-aware’ texts.”

The automation works on all channels with the same script, which simplifies the implementation of this type of service.

As for the future, CommBox plans to increase the speed at which it builds the script. At the moment all its interactions are written, but it is planning to introduce automated voice options.

The company, based in Kibbutz Glil Yam, central Israel, has deals with over 350 companies, from retail, to finance, and to healthcare – including Ikea, Adidas, Kimberly-Clark, AIG, Discount Bank and Clarks, the British shoe manufacturer. 

Subscribe to NoCamels weekly newsletter and get our top stories