Facebook Twitter Skype Linkedin Youtube Google Behance Dribbble
Menu
Rise of the Chatbots in Customer Experience
Dustin Brickwood

The year was 1966, the average cost of a new home in the US was around $14,200, Good Vibrations by The Beach Boys was playing on every radio and the first chatbot ever, ELIZA, was developed. Invented by prominent German Computer Scientist Joseph Weizenbaum, ELIZA’s inspiration derived from the 1950 “Computing Machinery and Intelligence” publication by mathematician, philosophist, and theoretical biologist Alan Turing. The publication defined machine intelligence as the ability to impersonate a human’s behaviour such that the recipient is convinced they are interacting with another human.  This instrument of measuring machine intelligence has stood the test of time and is still used today. Unlike ELIZA in 1966, modern machines are now passing the Turing Test that was created nearly 70 years ago.

I propose to consider the question, “Can machines think?”

Chatbot ELIZA

The rise of machine intelligence has brought the rise of chatbots, and as advancements in artificial intelligence and Natural Language Processing continues, the more difficult it will be to distinguish human from machine. The aim of this article is to provide readers with a general understanding of chatbots, what they are, their growing popularity, how they are typically built and their value proposition for enterprise organizations.

So what are they?

Chatbots are a computer program that impersonates human conversation in its natural format. They are able to communicate using sophisticated artificial intelligence techniques, such as Natural Language Processing, through text or voice interfaces. During the lifespan of the bot it will learn from previous interactions and will incrementally become more intelligent. Gradually understanding nuances in text or speech and handling more complexity throughout the conversation.

It’s not surprising that big businesses are looking to reap the benefits of these intelligent machines and with increased interest follows improved performance. Want to be among the businesses who get powerful insights and drive success with AI?

Contact us

Gartner, the research and advisory company predicts by 2020 the average person will have more chatbot conversationa day with a bot than they do with their spouse. Enterprises understand the benefits and advantages that chatbots can bring not only to their bottom-line but more importantly to their customers.

Value Proposition of Chatbots

It is important to understand that chatbots are being developed to complement and support humans, not replace them. The benefits of chatbots for enterprises are abundant but will never substitute human ingenuity and their passion. Below are a few use cases that highlight the potential for chatbots to drive customer satisfaction and sales.

24/7 Availability

Computer programs do not require sleep and do not mind working 24 hours, 7 days a week for below minimum wage…and people believe they’re smarter than humans. Customers now expect service at any time of day or night, they want their questions answered or their feedback  heard. Chatbots are being deployed to assist with this change in consumer expectations by providing an outlet for customers to engage and be serviced outside of business hours. Customers can now have their frequently asked questions answered, information requests delivered and brands can continue to provide exceptional customer service through automation, minimizing costs.

Bots are available 24 hours a day

Personalized Customer Experience

Conversing with someone who already knows your name, interaction history, and your product preferences will almost always deliver a better experience than interacting with someone who will repetitively ask for that same information. Chatbots have the capacity to integrate with customer relationship management systems (CRMs) and inventory management systems (IMS) to provide a heightened customer experience that is more efficient and engaging. Want to know when the new Google Home Hub becomes available in your city? A quick message to the Best Buy chatbot would respond with the correct information instantly. It even has the ability to ask you if you would like to pre-order the device, or browse a similar product instead. No longer are customers willing to look up a phone number, dial the number, go through pre-set menu options only to be told to stay on the phone while we transfer you to an available agent…cue instrumental music…and then to be put back on hold while the agent looks up that information for you.

Personalized Customer Experience

Analytics

Data is without a doubt the new oil and the more data you have, the better decision you can make. There is something inherently special about raw customer data, yet capturing the sentiment, range, and needs of a business’s customer base is challenging. Often businesses capture this type of data through surveys, complaint/feedback systems, or transcribed call recordings. Whereas transactional data is easy to capture, and although it can provide insights it cannot provide a 360 degree view of the customer. Using a chatbot enables the capturing of qualitative customer data that can be used to improve business processes, customer satisfaction or product selection. Every conversation is analyzed, comprehending the pain points of each customer and becoming increasingly aware of your customer’s true needs. With consumer demands evolving faster than ever, having real-time conversational (AI) data can ensure businesses are evolving with the customer faster, and allow strategies to be developed that follows the latest consumer trends.

How are these bots developed?

Chatbot development is primarily comprised of a channel interface, a Natural Language Understanding (NLU) or Natural Language Processing (NLP) engine, a bot platform that may allow 3rd party integrations, and a little creativity.

Channel Interfaces

A channel interface is simply the way in which users will interact with the bot, whether it’s through Facebook Messenger, SMS, website widgets, Slack or WhatsApp. Today, these channels are exploding with branded chatbots. With Messenger being used by over 1 billion users, it’s no surprise that there are over 300,000 bots on the platform serving customers every day.

Chatbot Channels

NLU/NLP Engines

This is the intelligence component of the chatbot. The advancements in artificial intelligence and research into varying techniques focused on natural language understanding/processing have been the driving force for the adoption of chatbots. The basic concept of these engines is that they have the ability to parse text or speech into structured data that can be used for further actions. Although there are many different implementations of NLU/NLP engines available, we will focus on intent classification machine learning which involves training the bot to understand intents by attributing them with multiple keywords and phrases that users are expected to input into the chatbot. The more the model is trained, the more it understands patterns and learns from them. It’s important to understand how these systems make sense of this data in order to develop a chatbot that correctly answers and guides users through its experience.  

Intents

An intent is a specific action that a user can invoke or the user’s intention. An example of intent would be if a user types “show me yesterday’s financial news”, the user’s intent is to see a list of financial headlines from yesterday. In development, intents are given names, often a verb or a noun, for this particular case it could be “showNews”. Further, we would define a list of utterances or keywords that would be attached to the relevant intent. For instance, under our “showNews” intent we could have the following utterances:

  • “Show me yesterday’s financial news”
  • “What’s happening in the financial news”
  • “Tell me some news about finance”

When the engine parses the user’s text and detects words associated with the above utterances or training phrases, it will match it to the “showNews” intent and proceed with an action. The action could be to retrieve a list of financial headlines from a backend service and display that information to the user in a list or gallery format for their review. In the case where the intent is not matched, we would provide a fallback intent, or also called a default bot response that would be triggered and inform the user that their message was not understood.

Entities

Entities are properties that assist NLU/NLP engines in identifying and extracting meaningful data from natural language inputs, this can consist of a name, location or date. When the engine parses the user requests and identifies an entity, that value can then be used within the following request. For instance, if a user inputted “Show me yesterday’s financial news”, the word yesterday and financial would be considered entities. When we make a request to retrieve news to send back to the user we will utilize these values to query news headlines that are from yesterday and considered financial related.

Artificial conversational entities

 

The world of AI and machine learning can be a very complex place but the above concepts are an integral part in intent-classification based engines. Luckily for developers and businesses alike, there are a number of platforms dedicated to abstracting the difficulty of setting up these models. Below are some of the markets largest NLP players:

  • Dialogflow (Google)
  • LUIS (Microsoft)
  • Wit.ai (Facebook)
  • Amazon Lex
  • RASA (Open Source 🤙)

Integrations

A bot’s capabilities can be extended using third-party API’s which can be integrated into the bot. When the NLU/NLP engine has determined the user’s intent, the next step would be to request that information from some source and send it back to the user to fulfill their request. There are a tremendous amount of possibilities of what information or action you could provide to your users. If we want to create a weather bot that responds to weather and forecast related requests developers can simply leverage an existing weather API, or maybe we want to schedule an appointment using the Google Calendar API. The possibilities are only limited by imagination.

Natural Language Processing Engine (NLP) for Chatbots

Conclusion

The chatbot space continues to evolve and with businesses increasingly allocating resources to chatbot development it’s only a matter of time before each chatbot developed will be able to pass the Turing Test. It is the relentless pursuit to provide superior customer service to demanding consumers where chatbots will be most commonly utilized. However, the possibilities are endless and the agony of being stuck on hold with instrumental music is nearing its extinction.

 

Facebook Comment

Read also

All articles