Why Q&A engines form another step in human language evolution

And how a good engine can resolve so many of your support problems

Everyone knows Alexa. And Siri. And Google Assistant, too. They are, basically, Q&A voice assistants. But why even bother getting into this adventure?

Before we explore the customer service market trends, let’s take a step back and observe how people communicate, and how do they communicate with businesses. We will take a short detour from the main course — in order to return to it much stronger.

If you think about it, communication is driven by evolution – natural and man-made. We can start from hundreds of thousands – even millions – of years ago when language evolved in early humans. They could finally convey information across time and space, without the need for both parties to actually be watching the event at the same time.

But it’s still was very limited: both parties for the conversation have to be at the same location – or at least, at a shouting distance. The amount of information is limited by the person’s memory — and by the time the parties have to talk. The invention of writing took humans a step further: one could now store much more information, and have it persist through location and time: you could read the document long after it was written and in a different location altogether. The first document we know of, written in Cuneiform, is not a poem or a war story: it’s a list of tax amounts no human could possibly remember.

Next came paper, which has less weight, so people could transfer it to greater length. Printing was revolutionary in that it introduced a new factor into the communication game: information could now be spread to more people, in minimal cost of printing another copy. Totalitarian regimes have limited ownership on printing machines in the fear knowledge would be spread without their control. Telegraph has shortened the time to pass the information, and telephone – the ease at which anyone, and not only professional Morse clerks, could deliver it. And of course – how can we forget -the Internet, allowing anyone to talk, watch and play with anyone on the globe.

So every time a new communication technology was invented and adopted, several things happened:

  1. The amount of data that could be transferred increased.
  2. The time and cost to transfer a fixed amount of data was reduced
  3. The number of endpoints to which the information is delivered increased

One cannot underestimate how profound was the change to people’s lives, on every new communication improvement.

Now, with the increase of computational power and machine learning models, we are at the beginning of another leap – with one new characteristic never imagined before. On all of the leaps before, knowledge was exchanged between two humans. For the first time, we can omit one side of the line, and put a bot instead. These bots – in chat and voice – can now stand as one endpoint of the conversation, and transfer information for millions of correspondents fast and cheap, as we would expect from a person. Now you can still think of print, radio, and TV be a form of such knowledge transfer for the masses, but here, finally, this form of mass communication could be bi-directional.

For businesses, such an ability can have a huge impact. A chatbot can understand the user’s intent and provide exactly the information the user needs. No more a single human agent needs to talk with a single customer. A bot can converse simultaneously with millions, at almost zero cost per additional user.

The cost of a customer service center

Calculating your customer service center cost is fairly simple, but it varies between industries. According to the Bureau of Labor Statistics, the 2018 median pay for customer service representatives was $33,750 for people with only a high school diploma, with an hourly rate of $16.23. Assuming an agent can have at least 5 calls per hour, a service call would be around $3. When products are more complex they demand a much more skilled agent, and in some industries, a call cost can start at $30 per Hello. A customer service center also carries hidden costs. According to research conducted by The Quality Assurance & Training Connection (QATS), the annual turnover rate in USA call centers is 30-45% with a cost of over $6000 for replacing an agent making $12 per hour. It means that in a 100 agent call center, only 70 agents are qualified to attend customer service calls.

According to Drift, the main benefit of chatbots is 24/7 service following by a quick response to users. Answering a simple question came “only” in third place. This is aligning directly with a survey from 2013 showed customers were waiting on hold on average for a total of 13 hours a year.

Customer service bot case studies

In this section, we will explore several chatbots and asses their impact. The first chatbot is called AVA that was created for Adobe and it was built on the IBM Watson chatbot platform. The main purpose of AVA was to be the front line customer service interaction solution for online communication. The main impact of AVA was response time reduction. From some interactions that took Adobe a day and a half to response were reduced into several minutes (by %99). Adobe stated that AVA is now:

  • Supports 30,000+ conversations per month
  • Recognizes 40 distinct use cases to quickly resolve easy requests
  • Cuts resolution from 38 hours to 5.4 minutes for most inquiries
  • Cuts cost per case from $15-$200 to $1

Such an improvement has a direct effect on increasing customer satisfaction and reducing customer retention. It is no secret Millenials are driving the industry and represent the major spending power these days. With modern-day standards, Millennials are much more sensitive to customer service quality and are willing to spend more on a product for better customer experience. In this post, we will not go deep down into the details, but a great blog post by Nextia has put together Customer Service Statistics and Trends for 2019.

The second use case we would like to present is TOBi – A chatbot by Vodafone. TOBi started as a simple Q&A chatbot and overtime got improved with new skills. In its second phase, TOBi got connected to Vodafone’s backend data center, in order to answer questions relating to the user’s account information. Today TOBi is taking care of 70% of customer’s queries in several areas – among which one could find technical questions, buying a new phone and more.

Jon Davies, head of the digital at Vodafone, shared what he has learned from building TOBi. In order to build a good chatbot you need to follow 4 steps:

  1. Gain trust of your customer – When TOBi starts talking, he presents himself as a chatbot. That sets expectations and trust – if he hadn’t, the user will eventually know it’s a chatbot in front of her. You do not want to get into this frustrating moment after the conversation has started.
  2. Build confidence – Improve your Chatbot skills gradually. Show users your chatbot is good at a narrow, small task, before adding more skills to it.
  3. Give your chatbot a personality – It is much more fun when you have a character and a story behind your chatbot. It is much more human-like. That’s what people want!
  4. Provide human agent support – in fact, these days chatbots are still very limited. They won’t be able to understand your customer as a real human would. Therefore, providing a way to have access to a human customer service agent is critical.

Plan the right customer service bot

With all that, as we wrote in previous blog posts, chatbots still suffer from many technological limitations. So before going into building a chatbot you need to carefully plan your path. Here’s a shortlist of what we usually recommend to our customers.

– Choose one problem, with a well-defined flow (or a tree of flows) that your chatbot will be solving at first

– Narrow the flow even further by using buttons and other pre-defined UI controls. This will introduce fewer cases for the developers to deal with.

– Check who is the audience of this chatbot. What personality do they expect from it?Once you have a plan, use numbers to compare: what is the cost of your customer service center, and what’s the cost of building a bot and operating it? In how much time will the organization receive its Return On Investment?


In summary, chatbots can be very cost-effective — if you choose the problem right, build them with the right platform and calculate the ROI correctly. And to add a nice touch to your efforts, you get to act as a player in a much bigger and dramatic human evolutional step.

Share this post

Share on facebook
Share on twitter
Share on linkedin