The Hub

Where we talk about ideas, events, cases & more
2 May 2017

Brand commerce & chatbots: 4 challenges for marketers

Customer centricity, marketing technology, customer experience, data,… More and more, marketers are realizing that these concepts are key in modern marketing. What it actually boils down to is “brand commerce”: Integrating brand building and sales activation with a focus on seamless and brand consistent customer experiences by making efficient and effective use of both data strategy and marketing technology. Chatbots are definitely one of the enablers for companies to reach this ultimate state of brand maturity.
The development of a chatbot is a hot topic on the agenda of numerous companies these days and in our Yearly Marketing Survey 2017, 5% of Belgian marketers say that they are already using a chatbot. But even the most eager of businesses should take a few aspects into account in order to ensure that their chatbot truly adds value for the customer.

1. A chatbot that speaks the language of your customers

It must have sounded like science fiction a few years ago, but it’s true: machines can read and reproduce human languages. You actually use them every day – just think of Google Translate, Apple’s Siri or Amazon’s Alexa - and the same principle is applied in quite a number of chatbots these days. It’s called Natural Language Processing (NLP).

Let’s start by a quick introduction into how chatbots actually work:

A basic chatbot usually has three main variables: user inputs, intents and actions. User input is anything that the user says to the chatbot (e.g.: I want to order pizza, I want to eat a pizza, get me a pizza now,…). These are all different user inputs for the same intent (order pizza). Once the intent is recognized, a pre-programmed answer can be given (Sure! What pizza would you like?).

The magic happens when the chatbot is able to understand more than just the user inputs that the marketer briefed to the developer. Thanks to NLP, the chatbot will also understand inputs that you haven't specifically provided (e.g.: I am hungry, can I buy a pizza?)

The issue? It works decently in English, but a large number of other languages are not yet flawless – just like with Google Translate or Siri. On top of that, language is highly ambiguous and often relies on subtle differences, making it rather complex for a machine to understand the meaning of words.


2. A chatbot that understands the meaning and context of words

Companies that want to go one step further can deep dive into knowledge bases such as ConceptNet, Google Knowledge Graph or Microsoft Concept Graph. These types of databases use machine learning to create enormous semantic networks of associations between different concepts, thereby enabling the machine to understand the context of words. For example:

ConceptNet

Most of these knowledge graphs are multilingual and are able to bypass the ambiguity of languages by focusing on the relationship between words instead of their definition in the dictionary, hence the advantage over NLP. Of course this does not take cultural differences into account. Therefore companies will most likely develop different graphs – and thus slightly different chatbots - for Europe versus Asia.

At a certain point in time, all possible concepts will be explored and companies will be able to re-use existing graphs to create their bot. The first companies who create their business model around these knowledge graphs have already popped up. Not only does this increase the efficiency of chatbots all over the world, but will also drastically reduce their time-to-market.


3. A chatbot that knows when to pass the conversation to a human 

Automation and technology have been hot topics in marketing these last few years, but there are things that humans do that machines simply can’t replicate (yet?). Think about empathy and creativity.

Chatbots aren’t the type of technology that can be developed, implemented and ignored. Instead, they must be followed up closely. When is it time to pause the machine and let a human take over the conversation?

  • Imagine that your chatbot is having a conversation with a very dissatisfied, but large and important, customer. Would you trust your chatbot to take the customer’s emotions and context into account or would you prefer a customer service officer to take over?
  • Or imagine that your customer shows a subtle interest in an upsell or cross-sell product in the conversation. Will your chatbot be able to recognize and use different sales techniques effectively?

Therefore, it’s also important to think about your chatbot’s ability to perform sentiment and context analysis during the conversation or to read signals of potential upsell or cross-sell possibilities and define rules for passing the conversation to an employee.

4. A chatbot that develops itself, instead of having to be developed 

The complexity and time-to-market of one single chatbot that can do and answer everything should not be underestimated. Instead, is it not better to start small by defining one clear objective and using machine learning functionalities to grow your chatbot step by step?

Machine learning gives chatbots the ability to improve their understanding capabilities and answers without human intervention. While your chatbot may not be first in class in the beginning, it will develop itself into an ultra-intelligent assistant and point of contact for your customers.

Let’s say that you are envisioning a customer service chatbot for your brand. Instead of specifying all possible intents, user inputs and actions yourself, a chatbot would be able to develop certain answers itself based on data transcripts from previous customer service conversations.


It is clear, we have a very interesting future ahead of us where technology will help us to proactively provide customers with what they need, when they need it. But marketers, let’s think things through and prepare for chatbots that solve real customer needs, instead of being mere Q&A assistants. Let’s put our data and marketing technology to good use, all in favour of customer experience.

 

References:

Vlerick Business School - Artificial Intelligence: the new arms race?
Chatbot's Life - Developing a chatbot? Learn the difference between AI, machine learning and NLP
O'Reilly Media - Using AI to build a comprehensive database of knowledge,
IBM developerWorks - 10 Steps to train an effective chatbot and its machine learning models

GettyImages-85184293_full-558069-edited.jpg