From The Ashes Of Chatbots
Recently, Facebook announced the demise of M, a chatbot assistant that once promised to be the future of conversational interfaces. First introduced in a limited public beta in August of 2015, the service was free to use and featured nearly limitless offerings.
M’s use of a human-trained AI allowed it to go well beyond automated conversation trees, and into the realm of sending a parrot to your friend’s office. It promised delight and awe, and a handful of competitors emerged, which I tested rigorously.
An Uber co-founder created Operator, a more limited commerce experience that raised 25 million dollars, and then turned off US availability to supposedly focus on the market in China. But the CEO has gone dark on the web, and there’s no indication that Operator is even still active.
GoButler was another worthy alternative that raised eight million dollars and was sold for scrap so quickly that some Kickstarter backers who used the service were left up a creek without a paddle (me, and someone else who spent $150 on an alarm clock rug).
There were a few others that disappeared quickly, too. All of which to say, the future of free conversational chatbots that help you shop on the internet and help you avoid phone calls, is mostly dead. (Magic and Fin are still around if you want to spend money though.)
Mostly dead is hard for me to type out because, in 2015, I believed chatbots were the future:
This is going to change everything. It’s going to improve online shopping — turning it into a frictionless and much safer process. It’s going to change how we work. It’s going to reduce the number of phone calls we make, and the amount of time we waste. We’ll teach the machines to do the tasks we hate, and our lives will be so much better as a result.
The problem with that hypothesis was three-fold:
- It’s really expensive to have free personal concierges standing by to handle a wide variety of requests. Even if you take a cut of each purchase, it’s still difficult and hard to scale properly. There’s not enough focus to be effective (which is why we’ve seen niche bots like Service and Rex have more success).
- Frictionless commerce emerged without needing a middleman. Apple Pay, Amazon Pay, and Google Pay all exist now to simplify shopping experiences. And if you’re really concerned about security, there’s a handful of one-use credit cards available — Privacy, Final, SudoPay, and even some actual banks too.
- People don’t mind handling small tasks themselves, especially when that provides more certainty and efficiency. Most shoppers enjoy the process of shopping, and the rest of us don’t care enough to work with an assistant.
However, where conversational commerce services failed, Alexa succeeded. You can now order more paper towels with a little Wi-Fi connected hockey puck on your counter. I was wrong – written chatbots aren’t the future. Vocal conversational interfaces are the future.
Perhaps “future” is a stretch. There are already tens of millions Alexa-enabled devices in use now, which is excluding Google’s & Apple’s competition.
But most of these devices are currently being used for rudimentary tasks like playing music, setting timers, and reading the news. Most people aren’t asking Alexa to warm up their car or send flowers to loved ones. That might be several years down the line.
To see where we’re heading, look no further than Betaworks’ voicecamp, a mini startup accelerator for voice-based interactions. They incubated tools that promoted mental well being, handled household tasks, and expanded written publications into the audio space. Which brings me to another prediction: in the end, I think we’ll see computing end up like publishing – a flexible mixing of audio and visual mediums that will adapt to what we need.
If you’re cooking, you’ll use your voice assistant to respond to emails, learn about the world, and manage your life. After you wash your hands, you’ll move back to a screen. And as computers learn to understand our voices with more accuracy, the fluidity will grow. Perhaps your rug alarm clock will even learn to understand half-awake mumbling.