Facebook’s Messenger Bots Don’t Help people Ask The Right Questions
Yesterday, Facebook launched the Messenger Platform which includes the introduction of ‘Bots.’ These bots can be almost anything. The core idea is that a bot can allow a customer to have a conversational dialog with a business, instead of downloading an app or browsing a website.
The vision for bots is at once both profoundly futuristic and old. A bot with sophisticated artificial intelligence should be able to do almost anything we might otherwise need a robust application to do. And do it quicker. But what they’ve really done is brought us back to a time when we interacted with real people instead of opening up an app. This is an old interaction paradigm, brought to life again with 21st century technology. Today, many businesses have attempted to replace this fundamental interaction between real people with an assortment of websites and apps. Has life gotten better? In many cases yes. In some, simply more cumbersome.
The Messenger Platform has the ambitious goal of substituting apps with conversation threads for the large majority of tasks. And the reasons are obvious. How often have you wanted to do something as simple as check in for an upcoming flight only to find yourself downloading the airline’s iPhone or Android app, jumping through a few hoops to find the checkin tab and your flight, and finally ending up with another app on your phone that you’ll never use again? Most of us have dozens of apps on our phones we either don’t use, don’t want, or don’t remember we have. Facebook thinks one potential answer is to replace an app with a bot that can act as a virtual concierge.
The big problem with these Messenger bots today: They do a poor job of helping a person know what question to ask. The interactions they are meaning to replicate (conversation between two real people) are complex and riddled with subtlety. A real concierge knows how to help a person understand what their options are. A concierge is perceptive of human emotion. A concierge is usually the first to ask the questions, and can provide materials to help a person do a few things on their own. In order for these Messenger Bots to succeed, they have this very high bar to meet and they certainly haven’t met it yet.
The bots available so far are beta experiences at best. And it’s important to note that they are really attempts at replicating the kinds of things I might normally do inside an app or with a simple web search. Unfortunately, their ‘artificial intelligence’ doesn’t seem very intelligent at all. I asked Poncho a simple question: “Is it going to rain today?” Poncho’s response didn’t fill me with confidence: “Wet. Warm. Yuck.” Rain was not on the forecast by the way. It’s also quite a bit slower to type “What is the weather in San Francisco,” than to just look at my trusty weather app siting on my home screen.
And there is still the problem of knowing what questions to ask. If you ask the Wall Street Journal about “stocks,” that word is apparently outside of the vocabulary understood by the bot. The list of terms the bot understands is unclear, and quite small from the looks of things.
Opening up a shopping bot, in this case Spring, deals with the prompt problem by asking some specific questions up front in order to narrow in on what the shopper may want. But in practice, the questions didn’t get much closer to a real search that I found useful. When I did ask a more specific question, the bot simply replied “We got your note. We’ll be with you shortly.” I have yet to receive a response.
This interaction model provides very little room for discovery. Maybe I just want to walk up and down the aisle. Maybe I don’t know what I want. And even if I do, I may not know how to ask the right questions that will give me a good response. The thing that websites and apps do really well and bots do not: They allow people walk around and explore. They give people the freedom to window shop without committing to a specific query. They allow people to move fluidly between just browsing, and searching for specific things. The web even lets people jump from one business to another nearly instantaneously. The good ol’ hyperlink is still tough to beat.
Shop Spring: “We got your note. We’ll be with you shortly.”
If Facebook Messenger bots are going to do a better job at all these things (and this does seem to be Facebook’s intention), they will need to solve more of these problems. In all likelihood, the bots will get smarter and faster. And some of these problems will be solved. But it is also very likely that a bot will never solve every use case.
Like most things, a business will need to choose the right tools for their specific needs. And generally the answer will be a unique mix of all tools available (websites, apps, bots, people, retail, etc.). Artificial intelligence and conversational UIs offer incredible new tools that product designers can take advantage of. As the ‘intelligent’ part gets better, things will get even more interesting. But it may be a mistake to think a conversational UI is always better than a simple website or app.
