Chatbots, from company gadget to must-have?
World of Digits / Explore
Over the past decade, most companies have been centering their digital strategy around a mobile app. But according to Gartner 1, companies will start abandoning the native app.
Enterprises are investing heavily to maintain and upgrade their apps, but also to promote them and surround them with the necessary customer support. Down the road original ROI calculations might turn out to be disappointing.
Customers experience an app fatigue, ask yourself this: when was the last time you downloaded a new app? Most of your daily mobile activity is focused around a couple of key apps that fulfill your needs. Aside from an occasional download of a revolutionizing app going viral, you probably are trying to limit the number of apps on your phone.
Together with the mobile web, virtual assistants form a good contender to become the main driver of how we will interact with brands digitally ².
What makes conversational interfaces particularly future-proof is its potential to deliver a fast, natural and highly-personalized experience. Why not interact with brands the same way we do with friends and family around us? Get information or a command a service via text or voice in a blink.
“We will have a travel agent, life coach, health advisor, personal banker or fashion advisor in our pocket that provides us with a real-time, personalized service”Yoav Barel 3, CEO & founder of Chatbot Summit.
But before worldwide user adoption takes off some pieces of the puzzle still need to come together.
Let us have a look where the chatbot market is today, using the Gartner Hype Cycle 4(Figure 1). This theory – in a nutshell- describes the life cycle stages of a technology over time.
A couple of years ago a handful of innovative chatbots turned out a success story, providing a tech trigger. I would not take long for media to jump on it and start sharing spectacular stories to the mass. A big buzz would follow and expectations inflated. However that enthusiasm quickly faded away. Band-wagon companies with few in-house know-how started launching premature chatbots causing tons of customer frustration. What was supposed to be a fluid conversation turned out to be loops of rephrasing sentences in a different way. The excitement of conversing with a machine was replaced by disappointment and frustration. The market took a trip to the valley of disillusionment. Companies asked themselves: “Is this it? Should we throw in the towel?”.
In 2017 the technology surrounding chatbots was not yet as premature: bot-development platforms were less user-friendly than now, chatbot analytics more a black-box than a feedback loop,
the NLP engines less performant. A NLP, or Natural Language Processing, is crucial for your chatbot’s performance. NLP technology deciphers text-based user inputs and provides meaning to it, so your chatbot can respond appropriately. Without a NLP a user-input would be nothing more than a combination of letters, not words, let alone sentences. Quoting George Kassagbi 5: “A mechanical music box knows nothing about music theory, likewise, chatbot machinery knows nothing about language.”
But also the user was not quite ready for chatbots yet. Today there is still a public fear and distrust over chatbots: some see it as creepy AI-monsters that will know everything about us, others relate it to people losing their jobs.
“It’s a complicated trade-off. We want to feel that the digital assistant is capable, but we also don’t want to feel intimidated by it.” 6
Users are not particularly nice to chatbots and, understandably, have very few patience with it. You would not believe how many people start cursing all over the chat when not being understood. Or troll chatbots with the most weird questions.7
Not to forget chatbots have to learn how to deal with the complex behavior of human beings and their languages. That takes time and requires lots of training data for machine learning. A single user intent can be expressed in hundreds of different ways. The way people talk or write is the result of a life-time of environmental stimuli. Meaning that a chatbot needs to be build in a way it is able to interpret that kind of complexity.
For example, when I was working on a trilingual bot active in the retail banking industry our team discovered an interesting pattern. After many sessions of analyzing the conversation logs we observed an interesting cultural, linguistic difference: our French-native customer base had the tendency to explain their context and situation extensively in the beginning before asking their question. Dutch people on the other hand seem to be more direct and straight-to-the-point. The differences in text length of the text inputs were quite impressive. It is hard to foresee such learnings in advance.
Every day technology is improving, the market is growing and learnings are being shared. More and more people are discovering chatbots and learning how to interact with them, and vice versa. To put it in terms of the Hype Cycle, chatbots are climbing the slope of enlightenment and heading towards the plateau of productivity (See Figure 1)
Today a paradigm shift is in the making. We are in the transition phase from humans using tech to tech using humans.
To explain this have a look at the evolution of the role of chatbots in relation to humans (Figure 2).
In customer service call centers have ran the show for years. Later mail and especially (live) chat started taking over the party. In fact, when given the choice 70% of customers prefer chat over voice for customer support – according to a 2017 IBM research 9.
And they are pretty good at that, according to IBM they can take up to 80% of routine customer service questions for its share 10. Few bots produce such figures yet, but shares are growing. Amit Ben 8, Head of Technology of LogMeIn: “The next logical step would be a human-bot symbiosis. Bots and agents collaborating together, evolving together to serve the common goal – satisfying the customer”. For example, bots will ask an agent for help when they are stuck in a conversation but afterwards simply continue the conversation.
Agents on their end will consult bots for previous customer experiences, customer data, and many more. Think about it, the potential is huge… a human-bot tango !
In a next phase a reversal will take place from technology serving humans to humans serving technology. Humans will only focus on those cases that cannot be automatized or for which the customer demands human intervention. It is just a matter of time before an ecosystem of different bots will take care of customer interactions. Each bot will have its own domain or specialty and they will even interact between each other. With the 5G network in the pipe, connected devices and Internet of Things(IOT) will only be accelerated.
Should your company build a chatbot? Depends.
Should your company investigate building a bot? If you ask me, definitely!
How to start? A good start would be to define the user personas of your product or service and to map the customer journey for each persona (in case you have not done that yet). Zoom in on the pain points in the journey and think about how a digital assistant can cure them. Or take a different road and brainstorm about the opportunities that will wow your customers!
Stay tuned – in a next edition I will zoom in on the nuances of building a chatbot and share tips & tricks to tune your conversations.
written by Floris Van Cauwenberge
1 Abandoning the native app?
² How we interact with brands digitally
³ Video – Yoav Barel, Founder & CEO Chatbot Summit | The 2nd International Chatbot Summit | Opening Keynote |
4 Video – State of the Bots | Yoav Barel CEO & Founder Chatbot Summit | Tel Aviv Jan 31 2018
5 Artificial Intelligence is Bullshit
6 What do we want from bots
7 Why are bots not taking over the Internet?
8 Video – Chatbot Summit Tel Aviv 2018 Keynote | Amit Ben, Head of Technology, LogMeIn
9 The future of call centers is shaped by artificial intelleigence
10 How chatbots reduce customer service costs by 30%