From ordering takeout to answering frequently-asked questions, chatbots now complete a wide variety of tasks across user experiences — but they’re not quite the silver AI bullet that we were promised. Why? The technology behind chatbots still has some growing up to do. These early versions have come up subpar due to a number of shortcomings, which we’ll go over here.
While AI and natural language processing (NLP) continue to rapidly progress, most systems require training to detect specific keywords from user input and trigger an associated action. If the user enters a complex sentence with multiple keywords, the chatbot gets confused — trying to figure out which action to prioritize. Most bots do not yet possess solid linguistic or natural language learning capabilities, meaning they barely understand human beings. And if comprehending the way we speak is tough, grasping emotion is a distant robot dream.
Chatbots rarely retain the contextual information produced during the course of a conversation. They tend to forget what the user previously mentioned and don’t apply a filter to the rest of the exchange, so they end up asking the same question repeatedly. Users either stay patient since it is a machine or more often, become frustrated, and wind up seeking live support.
Many companies implemented a conversational UI because it was a hot trend but neglected to build the necessary backend plumbing. For bots to be more effective, they need to automatically pull relevant data from customer relationship management (CRM), billing, or transactional systems to complete user requests. Without this connectivity, customers encounter a shallow, disjointed user experience.
When companies launch a bot for the novelty of it — lacking a clear strategy or purpose — it’s a recipe for disaster. Often, where an intuitive graphical user interface would suffice, developers introduce a clunky chatbot and blame the technology instead of the flawed strategic decision.
Bots frequently fail to get out of the way when they are unable to resolve a customer’s request. When the digital assistant cannot understand a user’s question, a seamless and quick transfer to a real person is the ideal scenario — preventing an annoying experience. Unfortunately, the software used by support staff frequently differs from the chatbot platform, forcing customers to explain the whole situation again to the agent rather than the data porting over automatically.
All these failures, however, have led to evolution. The next generation of chatbots looks quite different from earlier iterations, and open source technology advances are accelerating innovation at breakneck speed.
Big tech companies like Microsoft and Amazon have elected to share their natural language processing developments with the wider community. Thus, developers are capitalizing on these building blocks to improve chatbot functionality.
For example, NLP is getting better at the contextual understanding of human language. Exact sentence structure matters less as machines learn to interpret the wide variance in speaking and writing styles.
So far, the biggest challenge for NLP has been limited training data, but with the availability of big data, systems can scour documents and classify them by topic — reducing the need for manual training. These tools are increasingly able to teach themselves without human interference. As a result of the growing focus on bots for industry-specific data, NLP platforms are gaining momentum in fields like medicine, automotive, and manufacturing.
Machines can now express themselves better as well. Chatbots don’t simply show boring buttons — instead, rich UI elements and interactive mini-apps are part of the experience, making them a more engaging and powerful tool.
Finally, despite early stumbles, conversational marketing appears to be here to stay. Experiments are leading to better experiences and more intelligent chatbots. With the growing popularity of IoT devices, zero UI, and voice-enabled bots are proliferating.
The huge technology strides in AI, NLP, and machine learning are giving birth to a new generation of conversational marketing bots that offer game-changing upgrades.
Next-generation chatbots remember and learn about customers from data available across devices, systems, channels, and interactions. It’s possible to deliver a personalized experience, such as geographically-relevant information based on customer location. A convergence of large data streams and intelligent systems are underneath the hood of these updated virtual assistants, making them a lot more useful. Bots no longer have to be reactive. They can actively reach out to customers and engage them with in-the-moment conversations that are individualized and highly purposeful.
Chatbots continue to expand the number of interactive, single-purpose mini-apps and rich media embeds they can contain. These apps add practical value and new functionality to the bot experience.
For example, imagine a consumer walks into a store and discovers a way to virtually try on a dress by chatting with a bot on the phone. The person is guided to stand in front of a smart mirror which takes a photo and the person receives a rendered image wearing that dress in the chatbot. There is also an option to share with friends to seek their opinion. Contextual user information combined with media-rich interfaces allows for potent marketing engagements such as this.
New virtual assistants are serving as true companions to human agents. They do the bulk of the talking to determine customer needs and then escalate to the human agent who can step in as needed.
Like so many emerging technologies, chatbots fell prey to the hype cycle. They were thrust into the spotlight and spread like wildfire, trying to run before they could walk. The resulting disappointments soured the reputation of virtual assistants and led to a downturn in popularity.
The story of the chatbot is still in its early chapters, though. Newer, better iterations are emerging to solve big problems for customers and businesses alike. With these recent advances, conversational interfaces will evolve to a point where they become more relevant and fluidly adaptive to devices and situations.
Although it may seem like the bottom of the ninth, the bases are most certainly loaded with a bright future ahead for the comeback bots.
Vivek Lakshman is VP Innovation at Pramati Technologies, a global startup incubator, and technology investor.
The post How have chatbots failed to deliver on promise, and how will this change? appeared first on ClickZ.Reblogged 1 year ago from www.clickz.com