Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Getting a Bot That Really Works: The Best Corporate Architectures for Conversational Interfaces

DZone's Guide to

Getting a Bot That Really Works: The Best Corporate Architectures for Conversational Interfaces

Explore the best corporate architectures for conversational interfaces and why getting a bot that actually works is so important.

· AI Zone ·
Free Resource

The most visionary programmers today dream of what a robot could do, just like their counterparts in 1976 dreamed of what personal computers could do. Read more on MistyRobotics.com and enter to win your own Misty. 

People today expect instantaneity. That's why messaging has become the most popular interface in the world; billions of messages are sent every hour! But while the public has shifted to this conversational era, companies and governments are still living in the past. So, when brands record a drastic increase in customer support requests, they are failing to provide a good experience for their customers and losing them. Let's not forget that poor customer service amounts to $1.6 trillion annual losses in the US only.

"There are many reasons for a brand to build a bot! The two most important are innovation and customer relations. Innovation generates convictions about what you want to do with chatbots, giving you the opportunity to rethink your customer or employee relationship. All in the goal of offering a more powerful service."
Christophe Tricot, AI expert and manager at Kynapse  ( @ctricot )

In a past article, we've talked about why chatbots are the only viable solution to scale the customer service of major corporations. Remember how brands today are struggling to handle the booming number of customer requests and because of bad customer care, are losing money and gaining a bad reputation? Well bots, by excelling at managing simple customer conversations, are definitely the best solution: they automate support while increasing productivity and customer experience quality.

But how do you build an efficient corporate chatbot to achieve those goals?

Don't Think Vertical

To get a better understanding, it helps to understand how the chatbot market has evolved in the past two years. In 2016, the global direction was that chatbots were intended to fully manage conversations from beginning to end without human intervention. Any chatbot unable to do so was discarded as inefficient. When working with customers, we used to identify the top 5 most talked about topics in our client's customer support. Then, we'd work on automating them from A to Z with a chatbot. In the end, you'd have a chatbot capable of managing 5 vertical use cases perfectly.

"The three most common pitfalls of bot building are not having any real use case to support, not involving the right people from the beginning, or creating a chatbot disconnected from the company's information system."
Christophe Tricot, AI expert and manager at Kynapse  ( @ctricot )

But reality struck back. Even if you can manage 5 topics flawlessly, the majority of people using the bot will talk about something outside of its scope. It doesn't matter if the bot can quickly provide an invoice or update your contact details; if you want to know about the latest product and the bot replies, "I'm sorry I don't know what you mean", the experience is a failure to the user. And since the bot is discarded as an inefficient interface, the customer will not come back to use it, even if the bot could this time help.

Introducing the Receptionist Pattern

The receptionist is a design pattern we've created that thinks horizontally and not vertically. For any corporation, the goal is to provide a good customer support experience that leaves customers satisfied. No more, no less. Therefore, including a chatbot should not limit the range of topics supported. On the contrary, it should cover every area.

The receptionist is a bot design pattern that positions the bot at the beginning of every single user input. Every question passes through the bot, which is then capable of understanding what the request refers to. That's the key; the chatbot understands every single request. However, that doesn't mean the chatbot needs to treat everything. After understanding the question, the chatbot can either:

  • handle it autonomously through a fully automated conversation

  • start the conversation to gather important information e.g. client number, email address, and then hand the conversation over to a human agent

  • dispatch the request to the correct service and hand over to a human agent.

Because this architecture is horizontal, any customer request is taken care of smoothly. It doesn't matter if it is by a human or a bot, as long as the experience is smooth.

But if the bot doesn't manage things autonomously, what's in it for companies?

We've discussed the return on investment of one of our customers, a major telecommunication company in France, in a past article. By using the receptionist pattern, our client greatly improved customer support satisfaction by reducing conversation duration by half as well as by reducing the rate of multiple transfers. The chatbot was also capable of autonomously resolving 20 percent of all conversations, boosting the brand's productivity.

Another one of our customers in the telecommunications field also implemented the receptionist pattern. After only weeks in production, the chatbot's heavy usage on both the website and mobile app reduced the rate of aborted conversations from 15 percent to 0 percent.

Aborted conversations are when a customer starts a conversation, and never replies to the agent afterwards, most often because of the long waiting time. These are non-issues with the receptionist pattern! First, because there is no waiting time, users are less likely to drop the conversation. Second, the chatbot does all the onboarding, so if the user doesn't engage in the conversation, the chat will never be transferred to a human agent. Therefore, agents are not disrupted nor do they waste time with pointless requests, and the entire company's service quality increases, along with customer satisfaction.

And it's not just us!

"We've built a GDPR chatbot capable of informing users on the new rules and regulations of data protection. Users love it because they can quickly make sense of a complex law. That's why bots are great." 
Christophe Tricot, AI expert and manager at Kynapse  ( @ctricot )

So whether it's in customer service, marketing or sales, chatbots are becoming more and more sophisticated.

Robot Development Platforms: What the heck is ROS and are there any frameworks to make coding a robot easier? Read more on MistyRobotics.com

Topics:
chatbots ,conversational ai ,artificial intelligence ,customer service ,customer support ,nlp ,conversational bots ,conversational inferfaces ,machine learning

Published at DZone with permission of

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}