Section 508

February 23rd, 2017

Conversational Interfaces Are Falling Short but…

People want to chat and message with brands. There is absolutely no question about this. The ability to simply look up a restaurant on Facebook and ask if you can have a table for four people and get a direct response is great customer experience. The same is true if you have a question about your flight booking, the delivery time of a product, the source of a product or a customer complaint. We expect instant answers.

Worth noting here is that chat and messenger services are where people spend more and more of their time (10 minutes per day on average with 100% growth vs. 2016 according to eMarketer).

Also, 451 Research reports that “76% of customers prefer digital channels to communicate with businesses, and roughly 40% choose to chat with a business through SMS, social media or Facebook messenger, conversational technologies”.

These are all conversational interfaces, a messaging/chat interface where customers can engage with a brand. Why then are so few people using Facebook Messenger and messenger chatbots so far?

Why People Do Not Use Facebook Messenger or Chatbots

Michael Bayle, Head of Mobile at Amadeus, provided inspiring insights on this topic in his recent presentation ‘RU SIRI? US’. The presentation was intentionally cheekily titled to reflect his opinion that we have a long way to go to satisfactorily enable conversational commerce, via voice or thumb, but millennials’ thirst for such interfaces warrant our collective investment into this paradigm.

We decided to investigate how companies are performing today and analysed over one hundred well-known brands’ presence on Facebook. We found that:

  1. 9 out of 10 major brands don’t offer live chat or chatbots on Facebook Messenger which means that consumers don’t know when to expect what from a brand
  2. Most of the 50,000+ existing Messenger interfaces offer a really poor customer experience if they work at all (we will get back to this later)
  3. Even some of the more sophisticated brands failed to solve a real problem for the customer

In conclusion, the conversational interfaces currently offered don’t live up to customers’ expectations and do more damage than good to the brands.

What about WeChat?

One of the reasons Facebook and others went all in on conversational interfaces and chatbots last year was that WeChat was showing huge traction with integration of brands in the chat service.

However, when studying this in more detail, Nielsen Group concluded that conversational text interfaces did not play a key part.

“Our initial interest in WeChat was to better understand conversational text interfaces. … We did not find evidence for sophisticated natural-language understanding in WeChat. … Instead of a true conversational text interface, we discovered a system that warrants the interest of an evolutionary web scientist for the way in which it mimics the evolution of the mobile web — a world in which historic, simple interaction such as the numeric-menu selection and keyword-input coexist with more sophisticated menu-based interfaces or GUIs.”

Essentially, WeChat was fulfilling the role of mobile websites for brands and very successfully so.

Read more about WeChat in this excellent usability study.

What Are Some Good Examples of Conversational Interfaces in the Western World?


My favourite example is Starwood Hotels which was one of the first brands to truly embrace chat/instant messaging with guests in 2015. They simply let customers know that they could add the hotel property they stay at in WhatsApp and use it to communicate with the reception. Without any APIs or automation, they implemented a service that customers embraced and that Marriott (Starwood’s new parent company) later expanded across the group.

And Marriott doesn’t have a chatbot. Rather than offering a sub-standard experience they decided to staff Facebook Messenger with real people 24/7 and the result is probably as good as it gets.

Marriott chatbot



The travel booking service is one of the greatest examples of natural language processing combined with menu interface. Search for “Flight BCN to CDG 25/2 in the morning” and you will get the best price in seconds.

Hipmunk chatbot



Swelly, the crowdsourcing platform for decision-making, serves as a great example because they’ve nicely supplemented their app with Facebook Messenger with two beautifully implemented use cases. Vote or ask people for help to make a decision. It really is excellent and growth in active users is proving the investment worthwhile.
Read Swelly’s story about how the service evolved here.

Swelly chatbot



Sometimes the problem is not the most obvious one. KLM realized that what customers needed foremost was help along the journey, so they send the boarding pass and updates on your flight status through Facebook Messenger. Simple and yet so useful.

And yes, there are some other fun chatbots like Poncho (weather), Nerdify and Meditate but you will most likely use them once and then never again.

KLM chatbot


Who Got Conversational Interfaces Wrong?

Pretty much everyone else. Out of the 100 brands we tested, 95 either failed to respond at all within the first hour or provided a very strange response. And if brands don’t get the conversational interface right, someone else will own it as per Tom Goodwin’s article ‘The Battle is for the Customer Interface’.

Here’s British Airways from whom I still have not received any response:

British Airways chatbot


And Burberry, who showed leadership in conversational content last year during fashion week, seems to have lost their touch. Don’t tell me to ask questions if you cannot answer the simplest ones:

Burberry chatbot


Getting Conversational Interfaces Right Is Really Complicated

Setting up an automated conversational interface is easy. There are tools available that will do it in an hour. Building a great experience is however very, very hard. Just look at how Apple struggles with Siri, Google with Google Home, … And they have all the resources in the world.

Natural language is extremely hard when humans interact with machines as if they were humans:
Human: Set a reminder for tomorrow at 7 pm.
Computer: OK. What do you want to be reminded about?
Human: Actually, I changed my mind. Set the reminder for 7 am instead that I need buy a cake on my way to the office.
Computer: Reminder set for 7 pm to “Actually, I changed my mind. Set the reminder for 7 am instead that I need buy a cake on my way to the office.”

It gets even more complicated when multiple people are taking at the same time, but this doesn’t mean that we shouldn’t be trying. Now is the time to start learning and testing because in the near future the challenges with natural language will be solved.

What Are the Benefits of Conversational Interfaces?

As stated earlier, customers want to communicate using chat, messaging and voice. Many are even willing to adapt their communication to machines as we learned from SMS services a few years back.

Once we can meet customer expectations in terms of accuracy and speed of communication, and solving customer problems, the transition to conversational interfaces will go fast. It will allow businesses to serve customers better with a “human” touch anytime and anywhere and free up the time of employees to focus on the irregular requests and improving the service. More importantly, computers will be able to personalize and solve problems faster leveraging data, usage behaviour, machine learning and other insights. Overall, we will eventually see huge improvements versus communicating with “regular” humans.

What Does It Take to Be Successful?

Firstly, it’s about identifying simple problems that can be solved and that truly deliver value to the users. Secondly, be clear to the customer about what they can and cannot do, and offer service from an actual person or take them to the website as soon as it becomes clear their expectations cannot be met.

The best examples we found were by start-ups. This is probably due to the fact that they don’t have any preconceptions on what problem they are solving. Instead they work on figuring out what the customers want and need and test until they get it right. Just like the Swelly story referenced earlier.

The starting point can be as simple as telling customers when your team is available to help them and where to go when you’re not.

Keep it simple!

Magnus Jern
Chief Innovation Officer

DMI has 10 years’ experience of conversational interfaces (good and bad) including working with IBM Watson, Google Natural Language API, Amazon Machine Learning, Microsoft Bot Builder, Pandorabot and Facebook Bots for Messenger. Contact us for more information or human conversation.

Also, don’t forget to sign up for our upcoming webinar titled ‘Talkin’ About Bots: The Emerging, Mobile-Friendly App Alternative’ on March 1st at 11 am EST! A great opportunity for anyone who wants to understand conversation as a platform and get strategies for incorporating bots in their applications ecosystem. RSVP here.

Tags: chatbots conversational interfaces customer experience Facebook Messenger

Connect with us

Job Openings

Want to be part of our growing team?

View More
Work with us

Learn how DMI can help you grow, or launch your business.

Get In Touch

See all of our locations around the world

View Locations