One of my biggest pet peeves of working in conversational AI is hearing other folks referring to my products as “chatbots.”
I liken that to someone referring to a piece of Picasso’s art as a “drawing” because as anyone that has created a successful Conversational AI experience knows, there’s a unique art and process that goes into it that most “chatbots” just don’t have.
Chatbots by nature are fairly simple and basic. A user says something and the chatbot responds the best it can. From experience, I’m sure we all know that far too often that response resembles something along…
Over time, we humans have become used to venting and going on long, pointless rants when we’re frustrated. For the most part, this was never an issue as there was always another human at the other end of the phone call or chat session that was ready and willing to weed through the meaningless details of these rants and determine what we actually need.
In 2019, however, it is becoming increasingly common to see an AI-based tool such as a chatbot or a voice assistant at the other end of these conversations. This poses a huge problem as those meaningless…
If you’ve spent time scouring through conversation logs in Dashbot or Chatbase you’ve likely seen some frustrating, head-scratching conversations where your chatbot ultimately had the answer the user was looking for but never ended up serving it to the user.
Not only is this use case a huge pain but it can be a detriment to the long-term success of your chatbot if not accounted for. This issue is even more rampant when dealing with conversational topics that are more confusing for your chatbot’s users.
Hopefully, at this point in your chatbot journey, you’ve come to the realization that trying to predict user utterances for a given intent is about as hard as predicting the weather in San Francisco. Note the ‘about as hard’ as I’m sure there’s a lot of San Franciscans out there that would love to debate me on that.
San Francisco weather aside, if you’re pinning the success of your chatbot on predicting the endless options of utterances that your chatbot users could come up with, you’re bound for failure. …
In the early days of the chatbot, we were all treated to the concept of a generic one-size-fits-all chatbot, and oh what a treat that was. It was a pleasant surprise if the chatbot could even refer to the user by their name let alone leverage any meaningful data on the user.
In the end, we all became frustrated by drawn-out chatbot conversations that ultimately provided little value, and thus our pessimism towards the chatbot industry grew. …
If you’re looking for basic beginner-level chatbot UX design strategies, then I recommend you check out this article I wrote in 2018 on the ‘6 Chatbot UX Design ‘Must-haves’ for 2018.’ This piece you’re reading is an expansion on the chatbot UX design principles I covered last year and is for those looking to get their feet wet in advanced chatbot UX design concepts.
If you think most users will use your chatbot as expected this is clearly your first rodeo. Not only will user conversations surprise you but you will likely find yourself dumbfounded by many of them.
To expect a chatbot conversation to go as planned is the literal equivalent to meeting a stranger and expecting to know exactly how that conversation will unfold ahead of time. Plain and simple, it’s nearly impossible to predict chatbot user behavior.
Let’s be honest. When building your chatbot, you were so thrilled with the user experience that you expected users to flock to it in droves once it went live. Now you’re three months in post-live and practically banging your head against the wall when analyzing the user data.
Not only are users not lining up to use your chatbot as you’d hoped, but the numbers are well below the benchmarks you’ve set. Don’t worry; you’re not alone. Several inherent obstacles are a hindrance to virtually all chatbot user adoption, but they can be overcome. …
The 80/20 rule as it pertains to chatbots or support automation, in general, is the belief that 80% of your issues come from 20% of the potential root causes. This 20% of potential causes make up what we would typically refer to as frequently asked questions or FAQs for short.
Based on this, eliminating this 20% of potential causes would prevent 80% of the issues that come up. In short, FAQs can have a profound impact on the overall number of questions your support team has to handle.
One of the most significant challenges for those trying to offer service and support is being able to relay the proper internal knowledge to customers who need it promptly. As organizations and their product and service offerings grow, the amount of support-related information continuously expands. As this scope of knowledge and information broadens the gap between a tenured, experienced support agent and a new hire grows tremendously.
AI-powered automation can help bridge this gap and make all of your organization’s support and service-related knowledge accessible to your customers at the click of a button. …