Cat’s out of the bag: it may have been a long while since you’ve last conversed with a human customer service agent. More and more companies have turned to digital assistants to address the needs of customer inquiries. With messaging being the preferred method of communication, chatbots have taken over online facing services, due to their simplicity, affordability, and accuracy.
The following post will dive into how chatbots have evolved over the last decade and explore a few of the tools currently available to develop a digital assistant.
Chatbot Definition and Evolution
While they don’t need an introduction, chatbots are software
programmed to respond and provide information while operating behind a chat interface. They are useful for customer support, to direct toward informative resources or even to give information before transferring to a real agent. Many know these bots as the painfully automated answer machine whose
most common response is:
“Sorry, I didn’t catch that. Please be more specific.”
These first generations of chatbots can be seen as sophisticated decision
trees that often rely on If/Then logic to give you pre-determined answers. If you don’t use specific phrasing, the machines will not understand. While chatbots can be useful for a handful of situations, they often lead to more frustration than satisfaction from these interactions.More recently, you may have chatted with a smart chatbot, more commonly known as “ConversationalAI”. These chatbots are built on top of machine learning and deep neural networks to provide more sophisticated answers.Natural language is used to comprehend the intent of the person and a mix of dialog management and natural language generation is used to form a human-like exchange between computer and human.
Using neural networks, they can apply multi-turn judgment to direct the conversation. Hence, they are not limited to a strict script. Personality, humor, and wit are also sometimes applied to mimic a natural conversation rather than simply spitting out answers to questions.
These bots have the added advantage of learning on the fly; using each conversation to finesse their technique and get smarter with time.
While customers do not expect to have profound conversations with
automated customer service technologies, Conversational AIs have grown to limit the hassles and frustrations associated with traditional chatbots.
Now that we have a little understanding of what chatbots are and how they have evolved, let’s look into the various tools that can enable the
creation of one. Aside from coding one from scratch, many companies have ventured
into providing low-code environments to customize one’s own chatbot. Low-code bot builders allow an easy mean to make this emerging technology more accessible.
They rely on pre-configured libraries and drag-and-drop functionalities to ease development, fasten configuration and simplify deployment and testing.
For example, Motion AI is a chatbot builder that was acquired by Hubspot in late 2017. When launched, this tool was widely acclaimed for its simplicity and ease of use. It was long the preferred platform to create chatbots for apps such as Facebook Messenger and Slack.Motion AI enables brands to immediately engage with their consumers in a consistent and efficient way. With Hubspot’s CRM integration, this tool is now able to create more personalized messages thanks to the information already stored in the CRM.
IBM launched Watson Assistant in 2018, providing pre-configured libraries of inquiries and responses in the domain of customer care, banking, insurance, and more.Watson uses these repositories to find relevant content and walk the customer through a conversation. They also understand their own limitations, enabling them to stop and redirect the customer towards human staff when necessary.
More recently, Microsoft revealed Power Virtual Agents (PVA), a tool that allows anyone, even the least tech-savvy ones, to easily set up complex chatbots.
Similarly, to Motion AI and IBM Watson, PVA promises its users the ability to “create powerful chatbots—without the need for developers or data scientists”. What is particularly innovative is PVA’s ability to connect to other powerful Microsoft tools, such as Power Automate and Power BI.
With Power Automate (Microsoft workflow automation SaaS), users can enable actionable items from PVA conversation triggers. For example, if through PVA, a client asks to get a copy of his/her conversation, there can exist a trigger to email the entire chat to the necessary parties.Or, if a client mentions that their bank details have been compromised, a series of actions can be instantly launched to assist with the situation. Given the framework of your business’s needs, your chatbot can turn into a genuine virtual assistant that can enable actionable tasks with endless possibilities.Furthermore, conversations generated through PVA can gather many insightful details. When connected with Power BI (Microsoft business analytics SaaS), developers can easily visualize their bots’ limitations and conversation branches that were not previously considered.
This allows you to not only adapt to your changing clienteles but also forecast needs more accurately.
These tools are only a few examples of a long list of platforms allowing the democratization and accessibility to this technology. Each with its own sets of strengths and limitations.
Heading Into 2020
All in all, I’m impressed with the current state of chatbots technologies. The bots have greatly evolved over the past decade and, with the progress of machine learning, has been able to surprisingly camouflage themselves into our day-to-day.
As we are heading into a new decade, I’m convinced that with the emergence of no-code bot-builder tools, this technology is here to stay and prolifer in many more creative ways.
While the sophistication of AI chatbots is nowhere near the level of Ex Machina, I’m looking forward to seeing modern client-facing chat tools alleviate many of the burden and hassle faced by current customer-facing industries.The views expressed in this post are mine and do not necessarily represent EY’s position.