The future of Chatbots
Don’t give up on Chatbots just yet — advances in AI will revolutionize the interface.
Chatbots have already gotten a bad reputation. It may seem strange to say this, given that we are still clearly only in the very beginning of their development cycle, and that they haven’t yet really been widely implemented or found broad use beyond some flagship and still rather narrow ‘virtual assistants’, but it’s fair to say that most people are already disappointed by the ‘chatbot revolution’ that has been so loudly promised for the past few years.
Anecdotally, most people view chatbots as extremely limited in functionality, prone to mistakes, laggy, glitchy, lightweight pieces of software.
Part of this may be due to the nature of what is usually meant by the term chatbot. It isn’t always used as a catchall phrase for voice interface or chat interface implementation across a range of platforms such as web, mobile, apps or devices, but can be meant rather more narrowly to define a basic automated corporate presence on one of the major chat platforms such as Whatsapp, Facebook Messenger, LINE, WeChat or others. In this way of looking at it, the chatbot is to these platforms what a website was to the world wide web in its earliest days. Major corporations were setting up websites back then because they dimly understood that they probably should — but they weren’t very good yet.
Chatbots seem to be in a similar phase — either disappointing virtual assistant functionality, basic and inconsistently implemented website voice interfaces, or some kind of corporate ‘place holder’ account on one of the major chat platforms. It seems like we were promised so much more.
So, what’s the deal? Should we give up on voice and the chat interface? Was the revolution a bust? We think that is actually not the case at all and the best years of chatbots are still ahead of us. The reason of course is advancements in AI, which will allow for improved comprehension, systems that are constantly learning and improving, improved responsiveness, and the flexibility to understand and adapt to the messy and often difficult to understand real world.
NLP, or Natural Language Processing is the branch of AI seeing the most direct applicability to chatbots. And the state of the art in NLP is progressing rapidly. As quoted in a recent Forbes article, Salesforce chief scientist Richard Socher said: “NLP is going to be incredibly important for business — it is going to fundamentally change how we provide services, how we understand sales processes and how we do marketing.”
One source of disappointment in user interactions with virtual assistants is that they seem to be trained to only recognize specific combinations of words and if these magic words aren’t spoken, the system fails to understand the users intent. In this way, it is more like entering a specific command into a computer than having a conversation with a human being.
NLU, or Natural Language Understanding, is the subtopic of NLP that deals directly with this problem. Improvements in this area are allowing chatbots to extract user intent from natural language in spite of common errors like mispronounced or misspelled words.
In addition to understanding the words spoken, a flexible and responsive chatbot system must also be able to understand the ideas and concepts being expressed. Sentiment analysis is particularly important in our current social media driven world, in which brands exist online and are being constantly discussed by their customer base, fans and detractors.
NLG is another subfield of NLP, standing for Natural Language Generation. This enables personalized messaging to be dynamically created, based on the individual user — resulting in improved click through rates, conversion and customer satisfaction.
Another important concept coming out of NLP is aggregation. This is the ability of chatbots to learn from previous conversations to improve their performance. Since chatbots can carry out conversations at scale, this means they can be improving based on thousands of simultaneous conversations and improving their results in real time. The movement of AI processing capability to smartphones should accelerate this process.
Another AI driven advance to the technology that will improve performance is augmentation, which is the concept that the AI system can serve to assist a human by handling routine interactions and then seamlessly handing off to a human operator for more challenging or personalized interactions. This type of implementation could greatly expand the capacity of customer service personnel and sales professionals while improving customer experience and maintaining a personalized touch.
Even though chatbots are currently limited in their capabilities, it is clear that there is great promise in being able to reduce complex processes to simple conversations. So much of success in the mobile world is based on the ability of clever software developers to reduce or remove friction from experiences. If chat can similarly reduce friction, then it is virtually guaranteed success. Nobody needs instructions on how to have a normal conversation — it should just feel natural and thus frictionless.
The successful chatbot interfaces of the future may not be generalized virtual assistants but more domain specific implementations — or even a suite of specialized solutions that can work together behind the scenes and are accessed via a single voice interface to cover a wide range of use cases.
Specific industry solutions are being created for virtually every client facing business. Customer service, shopping assistants, travel assistants, tax and legal assistants. We look forward to seeing where this technology will go in the coming years and urge you not to give up on chatbots just yet. We’ve only scratched the surface.
By Angus Roven,
Neuromation Investor Relations Analyst
