Why Chatbots Are Failing Us

Image credit: iStockphoto/SIphotography

Chatbots are supposed to represent the future of the customer experience, yet so many consumers groan in frustration when one appears on their screen.

Several years into the evolution of chatbots, it’s hard to find anyone on the user experience side of the equation who will say anything good about them.

The line that they’d get better the more they trained and the more they learned is wearing thin, and for most consumers interacting with a chatbot can be a turnoff to further engagement through the website channel.

The prevailing sentiment is that the company doesn’t care enough about the interaction to provide a human, so the customer is consigned to a second-best option that rarely delivers.

Organizations seem to be learning this lesson, and research from Forrester show’s that apps are fast outpacing chatbots in terms of popularity.

“Even though a chatbot is often a channel within a channel like a website or an app, chatbots are falling behind native mobile apps,” a recent Forrester report says.

“Three years ago, the two were neck and neck at 17% and 19%, respectively, suggesting that customers used a chatbot about as often as an app. But since then, apps have surged by ten percentage points while chatbots have advanced by only five.” You can find more about the finding in this Forrester blog.

Consumer impatience

Forrester delivers a pithy diagnosis on chatbot implementation from Jon Altschuler, the senior director of creative services at LivePerson.

“In the early days, a lot of people thought ‘any idiot could put together a chatbot,’ and unfortunately they did,” Altschuler said.

Forrester estimates that 23% of online adults in the U.S. market communicate with businesses at least monthly through live chat or chatbots, whose popularity is particularly low in the financial services sector.

Consumers are impatient and have become used to technologies maturing rapidly. So when it comes to the evolution of chatbots, they are losing patience with the relatively sluggish speed at which they are developing.

“In the early days, a lot of people thought ‘any idiot could put together a chatbot,’ and unfortunately they did”

The Forrester report identifies some issues in the chatbot universe, and one of them is the tendency for the technology to have over-promised and under-delivered.

“Many of the people funding chatbot initiatives have expectations that are not realistic about what chatbots can do,” according to Sascha Wolter, chief advisor for UX/conversational AI at railway Deutsche Bahn.

Brian Smith, auto insurers USAA’s lead digital product manager for conversational AI, told Forrester that “today’s platforms have many limitations — but a big part of the problem is people not truly understanding what conversation design and AI are.”

Long term future

Chatbots, however, are not a lost cause, but they need to improve if they are to have a future in the longer term.

One problem today, according to Forrester, is that “making the chatbot primarily an IT project is a recipe for failure.”

“There are many tech providers offering chatbot platforms that are either deterministic or based on machine learning (ML) or use both,” the report says.

But the technology is still in the early stages. ML for natural language processing (NLP) has made impressive progress considering where it started, but overall, it has barely lifted off the tarmac.

And just like most of the rest of AI, it will be many years before it reaches cruising altitude.”

As with many developing technology areas, the solution lies in hiring someone from a profession that many people have not heard of yet — a conversation designer.

Design and not IT is the key to success, and the success of the chatbot is dependent on its level of conversational expertise, and for that, hiring an expert is required.

Another piece of advice is not to go straight to Artificial Intelligence. A well-designed and clearly mapped out decision tree is often the best place to start, and then other technologies can be implemented as the chatbots prove their worth.

A combination of a good decision tree and a smart conversation designer, Forrester says, should be the starting point for a chatbot project.

Because they deliver words, chatbots are entirely different from most other consumer-facing user experiences, which are predominantly visual.

Where you might clock on a drop-down menu to bring up a field you want on a website, you can say anything you like to a chatbot, and the parameters of what you say are not defined by the IT.

Chatbots aim to replicate a human interaction more closely than other digital interactions and to deliver that requires a totally dedicated design focus.

Another piece of advice is to understand what you want the chatbots to do clearly. Do you want them to deflect calls, build loyalty or sell more? At this point in their development, they probably can’t do all three.

Forrester also emphasizes the importance of analyzing voice logs as part of the ongoing evolution of the chatbot's conversational capability. It’s not enough to think that if you have embedded AI in the chatbot, it will get trained up without it.

Ultimately, the lesson is that chatbots are not set and forgotten. They need human-centric design, and once implemented, they need nurturing if they are to charm and not annoy the customers organizations are hoping to delight.

Lachlan Colquhoun is the Australia and New Zealand correspondent for CDOTrends and DigitalWorkforceTrends, and the editor of NextGen Connectivity. His fascination is with how businesses are reinventing themselves through digital technology and collaborating with others to become completely new organizations. You can reach him at [email protected].

Image credit: iStockphoto/SIphotography