Bots are hot in Silicon Valley.
Everyone’s obsessed with “conversation as a platform,” or the idea that you should be able to interact with a service with natural language, spoken or written, instead of in the traditional click-an-icon way you’d interact with a computer.
Google, Facebook, Microsoft, Amazon, and a slew of startups all rushing to find new ways to make it easier for users to connect to all the services they need through a single interface. Just this week, The Information reported that Apple plans to open up its virtual assistant, Siri, to third-party developers who want to integrate their services. We’ve entered the age of bots.
But with the explosion of new companies and initiatives, there are some of the challenges and lessons learned:
Less C3PO, more R2D2
Evernote founder and partner at VC firm General Catalyst, Phil Libin, believes that bots are the most exciting tech he’s seen since the iPhone and that they will rewrite the tech space like apps did in 2008.
Libin, who leads incubator within the firm to experiment and build bot companies and has already announced his first investment — a company called Begin — says that everything he’s seen has taught him that the “first law of chatbots’ ought to be that they shouldn’t try to deceive you into thinking you’re talking to a real person. If they do, people will just end up disappointed.
In Star Wars terms, bots should be more like R2D2 than C3PO: You talk to R2D2 naturally, but it doesn’t sound like a human in response, whereas C3PO does talk like a proper human despite being a robot.
“I think most bot developers have this C3P0 model stuck in their heads, where they’re trying to make bots speak like humans and give them jokes and personality, but I think that’s really misguided,” he said.
Beerud Sheth, CEO of a bot developer called Gupshup, agrees:
“A bot that pretends to have human-like natural language processing and AI capability raises the user’s expectations, which inevitably leads to disappointment,” he says. “A bot that sets clear expectations about its capabilities (and limitations) will have a better shot at engaging and delighting the customer.”
Be a guide, not a butler
That underscores another problem right now: Most bots aren’t that smart. Or, as Libin puts it, we’re still stuck in the “fart and flashlight” era of usefulness.
So, bots need to make it clear what their limits are and try to help people get as much use out of them as possible. In other words, they should be guides not butlers.
“A chat bot shouldn’t just be responsive; it should also guide the conversation in the ‘right’ direction,” Sheth says. Right at the beginning, the bot should introduce itself explaining its capabilities. If it can’t answer something, it should respond with alternative suggestions about what it can answer.
Don’t bother me
Especially in this early period, bots that send too many notifications are likely to get blocked.
“Chatbots shouldn’t proactively contact users unless they are providing some really valuable insight, otherwise, they come off as spammy,” says Rob Mob, CEO of a Talla, which makes intelligent assistants for business. “In general people prefer reactive bots, but proactivity is fine in cases where the information is useful.”
He added that the most surprising thing Talla has learned is that bots shouldn’t respond too quickly. People prefer a small delay, to make the interaction feel more like normal communication.
Although chatbots are “very weak” right now, Mikhail Naumov, president of the bot company Digital Genius, says that with time they will grow up and be able to learn more quickly and deeply.
“I think right now we are in the ‘Valley of Disillusionment’ part of the hype cycle,” Libin agrees. “The early stuff was all trivial and disappointing, so the people who were chasing trends are now sceptical. While people who are serious haven’t even begun to deliver what they’re working on. I think the people who are serious about them, it’s going to take another several months for even the early work to start coming out. It’s an exciting time.”