ChatDirect - Nieuwlandsedreef 141

After web, mobile & apps it's now the era of chatbots as part of the technological revolution. But how do these chatbots work and why will they be very important within customer service?

For those of you that don't know what a chatbot actually is and why they are rapidly taking over I can recommend this short Hubspot video:

“If you're not building a bot, you're already behind.”

Seven reasons why chatbots will play a crucial role within call centers:

1. Chatbots make no mistakes, never get tired and work 24x7x365

Let's start with the obvious reason: a lot of the tasks we execute on a daily basis can be done way better by computers. This for sure applies to call centre employees: 80% of the question coming in are easy to handle. Ideal for a chatbot available every hour of the day, 365 days per year, waiting there to help customers at scale. Thanks to chatbots live agents have more time to help customers with the more complex issues resulting in an increase of customer satisfaction with double figures.

2. Chatbots are way cheaper than call centre agents

Recruiting and training a live agent costs at least a couple of thousand euros (I even didn't take the salary into account here). This means chatbots are also cost effective for the small contact centres.

3. Chatbots have a very good memory

If built correctly and supported by a solid Customer Relationship Management (CRM) system, personalisation can add great value to chatbots. When the user is logged in the bot knows within milliseconds the user's history regarding orders, sales and previous contact moments. The conversation can be adjusted to this information creating a personalised user experience.

4. Pro-active customer service

By monitoring website behaviour realtime, bots can interfere even before a problem occurs. Powered by artificial intelligence and machine learning, situations can be recognised in which users tend to drop off because they can't find what they are looking for and get irritated. A chatbot can offer a helping hand at exactly the right moment.

5. Almost half of consumers already prefer talking to a chatbot instead of a service desk agent

Going through 20+ options of an interactive voice response (IVR) system and waiting in queue to be served is one of the biggest consumers' annoyances. In case it turns out the help desk employee doesn't have the needed knowledge and/or authorisation to help out, "disaster" is complete.

6. Messaging platforms like WhatsApp & Facebook Messenger already surpassed social media in usage

People like to communicate with businesses the way they do with friends and family. With messaging platforms like Messenger, WhatsApp, SMS and Telegram consumers can have frictionless conversations with companies because no new app needs to be installed from the app store.

7. Chatbots offer multi-channel communication

Imagine asking a question to a chatbot in the morning via Messenger. You don't have enough time to finish the conversation but as soon as you drive off to work your car assistant asks if you want to continue the conversation via voice chat.

How chatbots work: Natural Language Processing

A common mistake is copying existing structures from websites or IVR menus. Chatbots offer the opportunity to figure out what a customer needs via conversation. This is called Natural Language Processing (NLP).

Understanding intention, entities and context

Besides the intention the bot also needs to find out which entities apply and what the context is. Let's take an example at a bank: a customer, we call him Hank, wants more information about taking a second mortgage on his company building.

Intention: getting a mortgage. Entities: mortgage, company builden. Context: business.

Human language challenges

The fun thing about human language is there are many ways to say the same thing and this is where one of the biggest challenges awaits: program the bot in a way it can easily filter out intentions from a conversation and becomes smarter when more conversations have been handled.

Asking questions

Asking questions is an important method to identify the intention of the user. Hank can for example ask the chatbot for more information about a loan for his company. The bot could ask if Hank owns the company building and suggest an extra mortgage as one of the possible options.

The follow up proces

A chatbot is also very suitable to arrange a follow up step, in our example: setting up an appointment with a representative at the client's business, at home, at the local branch of the bank or via a video call.

A combination of free text input and fixed flows and quick replies often works the best. As long as the intention of the client is not known yet often free input is used and when de possible outcome is known fixed flows are the most efficient.

Machine Learning & Artificial Intelligence

When a computer has cognitive skills to solve problems and learn from circumstances and historical data we call it artificial intelligence (AI).

A chatbot for a bank website needs to be able to answer a lot of different questions and different types of clients might use different kind of lingo. With Machine Learning & AI chatbots can "get to know" their users and improve conversations over time.

Bots can for example detect when it can't answer certain questions and suggest to the human maintaining the bot to add content about a certain subject.

Besides that the system learns synonyms used for the same intentions, entities and contexts. This can be done by comparing the initial question that couldn't be answered, with the endpoint of the conversation.

I think a combination of man and machine is the best way to go to make chatbots smarter. The system suggests improvements and a human being approves, adjusts or rejects them. Especially for organisations starting with chatbots this approach is recommendable to stay in control and keep the overview.

Personalization is the key to success

When Hank is logged in better service can be provided because the bot knows exactly the financial situation of the user: does he already have a loan for his business, does he own the building, how often does a negative balance occur when looking at Hank's bank account(s)? Of course, extreme cautiousness needs to be applied when working with sensitive data like this. It must be crystal clear to the user that the conversation takes place in a secure, restricted environment. Besides that it's recommendable to simply ask something like "Is it ok if I pull your personal records from the system?".

Create an avatar and use humor

People seem to appreciate it enormously when a chatbot possesses human character traits and some visual resemblance. Most bots are given a female name. The two biggest banks in the Netherlands for example, ABN AMRO & ING, have called their digital assistants Anna and Inga.

An important human characteristic is having a sense of humour. Of course, a website related to funerals is not the perfect place to do this. The amount of humour that can/should be integrated depends on the kind of business. This aspect of a bot is important because it creates a feel good experience that will be associated with the brand.

Chatbots are no holy grails solving all problems

Of course, a chatbot is not a holy grail turning a poorly organised company with mediocre customer service into a well functioning one. The fundament, a solid Customer Relationship Management system, needs to be in place.

Tips for starting out with chatbots

The most important advice I can give is to start small and iterate step by step. Define a "minimal viable chatbot" to prevent building features that won't be used. Include end users in the development process of the bot early on, collect feedback constantly and use it to make your chatbot better.

The age of robots is now!

This article was first published in Dutch on Upstream.nl. For more great content about chatbots, voice, artificial intelligence and conversational marketing please join the Botrepreners community at https://www.facebook.com/botrepreneurs/.