Why most Chatbots suck
As I am in an Uber on my way to the Airport, I get stuck in traffic. There is no way I can make it in time for my flight. In a blink of an eye I decide to open my Travel app and start a chat with my travel agent. A couple of minutes later I am boarded on the next available flight and have paid for the extra costs. Technology is great, isn’t it…. But why do I need to type a question in a stupid app with a small keyboard if the system could have known I was delayed based on my context? And why didn’t the app alert me in advance to avoid that route? And what about connecting flights?
Artificial Intelligence and Robotics offer many exciting opportunities to enterprises, things that sometimes go beyond our imagination. There are so many cool technologies around that for most entrepreneurs it must feel like being in a huge candy store.
“Please give me that Robotic Process Automation-solution and oh, I would also like some Natural Language Processing and a chatbot or two.”
The requests for Proofs of Concept are piling up on the desks of people like me at high tech consulting companies like Accenture. But unfortunately, unlike candy, AI-components do not bring instant satisfaction beyond the first eye candy of having a robot answering your email or a chat. On their own each of these technologies will not bring anything more to an enterprise than a short sugar rush. And that is why most chatbots and other AI experiments currently suck.
The bigger picture
Do I state there is something wrong with AI Technologies like chatbots? No, certainly not. Independent of vendor and technology, I am convinced most AI initiatives can bring huge value to society as long as they are designed and implemented in a proper way and embedded in the processes and culture of an enterprise, taking into account the full customer journey from an end-to-end perspective. A Proof of Concept of one isolated technology tells us nothing about the value of a technology and how it can support and motivate people. As each of these technologies has its own strengths and challenges, combining them in an enterprise-wide AI Ecosystem is the only way to make them effective.
Chatbots are for example very strong in interpreting the intent and sentiment of a person and can - based on algorithms - respond to those in a proper way. They however have issues with more complex situations (beyond IF…. THEN… and simple expressions) and often do not adjust to the real context of a person. This can make the whole interaction forced and ‘artificial’, especially in situations where the chatbot is used as an alternative to a non-optimal web interface. Even worse are the situations where the chatbot – like annoying sales persons in a shop – bothers you with a “How can I help you?” when you browse an ecommerce website.
To make Chatbots and other AI Technologies effective, we really need to look at the bigger picture. Let’s go back to my initial example, the one that has been used by dozens of chatbot vendors to sell their technology. From a customer perspective I would like to be constantly informed, instructed and supported during my journey from A to B, taking into account my real-time context. From an enterprise perspective the technology should also avoid unnecessary human interaction with the customer as this costs time and money.
An Orchestrated Process
So, what would my customer journey look like in an orchestrated process that takes into account the full context of the customer.
As I am in an Uber on my way to the Airport, my phone alerts me of a traffic jam on the route to the Airport. It offers me two options:
- Adjusting my ride to include Public Transport, avoiding the traffic jam all together and arriving in time on the Airport for my flight. The App provides the address of the nearest train station and all I need to do is adjust my Uber ride to go to a different location.
- Rescheduling my flight and all connecting flights and continue my ride with the Uber. This option could for instance be more interesting if I am traveling with a group and/or with a lot of luggage. Based on my preference on for instance arrival time, budget and airline, the system automatically books a ticket for another flight.
My choice is recorded in the system and a case is started that guides me during my journey and handles all tasks. All I have to do is choose from the options it provides. If I like I can still type questions, or use my voice to do so, but this is only necessary in the rare occasion of an exception.
Viva La Evolution!
Building a customer experience like I described above, is not something you should try implementing with a big bang (although that would be cool). Organizations need time and attention to transform and embrace the combined hybrid workforce of expert employees and AI. Most enterprises have already started with implementing a Virtual Digital Workforce or Robotic Process Automation (RPA) and some are now moving towards the more advanced cognitive technologies. My advice to these organizations is to take one step at a time and transform the organization gradually into this agile, scalable context-driven hybrid workforce; ultimately offering customers end-to-end services that go beyond the possibilities of each individual component in the system. Try to avoid the pitfall of choosing technologies that look good in a Proof of Concept but fail to scale. Think in end-to-end storylines and find the technologies that allow you to support and orchestrate the underlying processes.
In the meantime, keep on reading my articles and don’t hesitate to contact me if you need inspiration or have a great idea that you want to share.
My articles are my own views and do not necessarily represent the views of my employer, Accenture.