With Current AI Tech, The Dumber The Chatbot, The Better The Results

by Tim Sloane 0


This opinion piece in Forbes, written by a Chatbot expert, suggests that it is a mistake to rely on current AI technology to cover broad areas of expertise. Instead focus on increasing convenience by making it easier for consumer’s to take action within a more limited knowledge domain:

“All-knowing bots that attempted to do everything from making jokes to ordering plane tickets rapidly veered into the valley of discontent because they couldn’t decode the user’s intent accurately in numerous use cases — their intent library just wasn’t thorough enough. These early bots tried to do too much and often failed.

Customers have a low tolerance for failure, and chatbots quickly gained a reputation similar to that of interactive voice response (IVR), another automated technology that promised streamlined communication and instead resulted in increased customer frustration. This perception of chatbots has been a massive blow to an industry that in fact can and already is streamlining communication.

Today’s best bots solve narrow problems. They don’t try to be human, they don’t do everything and they can’t hold a continuous conversation on most topics. What they can do, though, they do very well.

Apple’s recently released Business Chat is an excellent example of how bots and messaging can be used as a seamless substitute for web forms. At WWDC 2018, Apple demonstrated how Business Chat can collect information upfront from a user, such as the user’s login credentials, before handing the customer off to an agent. The chat screen uses conversational bubbles to make entering information as seamless as possible and eliminates the need for the customer to switch between browsers, apps or email.

This same format can be used for any web form — from payments to purchase decisions. Geico, for instance, uses a bot to ask users for details in order to select the correct insurance plan for the user’s home or car. The messaging format works well because it is incredibly mobile-friendly and it also leads seamlessly into agent conversations if need be.”

The article goes on to highlight the types of problems chatbots can solve well and suggests that it is important to Underpromise and Overdeliver:

“Bots today are highly adroit at solving problems that have a limited set of outcomes. The two primary types of problems bots can solve are issues of search and identification and issues that can be resolved through frequently asked questions (FAQ).

Search and identification problems consist of the user describing through a series of questions and answers what they’re looking for, and the bot surfacing a solution. For instance, multiple airlines have employed bots to help mobile customers find flights. 

Frequently asked questions are one of the best applications of bots because they deflect tickets from customer service. The Amazon customer service bot is a great example of a bot that does this well: After users give a bot enough information about their issue, the bot can suggest knowledge base articles that address the user’s problem and only escalate to a human agent if the article does not answer the user’s question. Like finding a flight, the bot is simply matching the customer’s needs to an outcome.

Today’s bots may not be as flashy as their predecessors, but they offer something better: They are simple, easy to use, reduce friction and function effectively 99% of the time. These are the bots that underpromise and overdeliver.”

Overview by Tim Sloane, VP, Payments Innovation at Mercator Advisory Group

Featured Content