Everyone seems to be jumping on the bot bandwagon. Chatbots are not only being touted as the end of apps, but also the next paradigm in human-computer interaction — and, if you believe the hype, the end of many customer service jobs, too. But bots are unlikely to live up to these outsized expectations anytime soon.
The fact is that the AI technology used to power chatbots simply isn’t mature enough to come close to replacing humans for anything but the most trivial tasks — the same ones that are already well-suited to apps. With expectations set so high, be prepared to see a lot of frustration and disappointment as reality sets in.
Conversational interfaces: To bot or not
We increasingly expect brands to adapt to our lifestyles and engage on our terms. That includes embracing messaging as a first-class communications channel. Conversational user interfaces and B2C messaging have the potential to radically evolve consumer interactions, but if brands rush to hand off their customer interactions to algorithms, they’ll be making a costly mistake.
Conversational interfaces are only as good as the communicator on the other end. If bots are only serving as text-based substitutes for automated phone systems, then we’re just providing consumers with another channel for frustration. They’ll learn to immediately type the chat equivalent of dialing “0” to get a human. Meanwhile, customer care executives will be disappointed when the promised savings and efficiencies fail to materialize.
This is unfortunate, because a negative consumer reaction to bots threatens to tarnish the perception of conversational interfaces and B2C messaging as a whole, similar to how 800 numbers have become associated with ineffective IVR systems and long hold times. The medium is the message, and if messaging a brand means providing a bot that can’t understand consumers, you are basically saying to consumers “we don’t care.”
What are bots made of?
Bots purport to be a paradigm shift in consumer communication, combining cutting-edge AI and deep learning with the ability to integrate into existing messaging platforms. But the AI powering these bots is already available in other implementations — and it just doesn’t work reliably yet.
With expectations set so high, be prepared to see a lot of frustration and disappointment as reality sets in.
We’ve all experienced this in the form of both virtual assistants (Siri, Cortana, etc.), as well as the ubiquitous and frustrating automated systems used in telephone customer service. While they have their uses, most of us are painfully aware of the limitations of both. Sure, bots eliminate the voice recognition and voice synthesis steps, but those aren’t where the real challenges lie.
Understanding (and speaking) the words isn’t hard; it’s understanding the meaning that’s the challenge.
The limits of robotic communication
It doesn’t take long to realize that these technologies can’t really comprehend what we’re saying. They primarily listen for keywords. While researchers have made significant progress in areas like natural language processing and context awareness, those advances don’t necessarily offer more practical utility in the implementations that are available today. In real-world use, even mildly complex sentences don’t register. Try asking Siri for the weather in a friend’s city. It doesn’t work, even if that friend’s address is in your contacts.
The second problem is access to information. Even if an AI algorithm can understand the question, it doesn’t necessarily have the answer. This information is often stored in massive legacy databases that are configured to handle specific queries. Bots aren’t plug and play, and making this information accessible will take significant engineering work.
The dirty secret: Bots are people, too
At present, the bots that are even remotely useful at non-trivial tasks are largely supported by humans behind the scenes. While the industry continues to promote bots as a panacea for everything from app development to reducing headcount, companies are quietly employing human labor to buy time while developers attempt to implement AI solutions that can work autonomously and scale as promised. The software is supposedly learning with every interaction, but whether or not these bots ever graduate to real utility remains to be seen.
Bots can respond to simple questions like balance inquiries, but they’re handing off the heavy lifting to people, as will be the case for the foreseeable future. In this respect, bots offer some utility — acting as a front-line receptionist for customer care professionals by regurgitating basic information, while triaging everything else.
As a tool in the customer engagement arsenal, bots have value, but right now, they don’t come close to living up to the hype. Nevertheless, the more information our automated apprentices have to learn from, the better. So brands that want to leverage bots tomorrow should begin by building rich archives of conversational data today.
Brands should proceed with caution
Even in this more realistic hybrid scenario, the devil is in the details of the implementation. Right now bots aren’t very smart and they can be incredibly slow — the worst of both worlds. Waiting for a response from a human is, if not enjoyable, at least understandable. But waiting for something that’s advertised as a computer seems absurd, and infinitely more frustrating.
Avoiding the disconcerting uncanny valley that emerges when machines try to mimic humans is a delicate balance. Bots will need to queue and distribute requests seamlessly to their corporeal counterparts without the conversation feeling schizophrenic. We don’t expect traditional apps to care or exhibit empathy, but if software acts human, then consumers are going to hold it to the same standard as humans.
By setting the bar so high, anything less than perfect execution stands to be a huge disappointment — and brands can’t afford to shirk responsibility for the customer experience by adding a “beta” label, regardless of the technology behind it.
Using messaging as a playground for experimental AI is a good way to alienate consumers at a time when meaningful connections are an increasingly scarce and valuable commodity. Hopefully bots evolve enough that human agents can be freed up to spend their time helping the people who need it most. In the meantime, brands would be wise to treat those interactions as precious opportunities for engagement rather than rushing to hand them over to bots.