The bots bubble has burst, but they’ll be back

Posted by
Rurik Bradbury
Head of Global Communications and Research
Wednesday, April 12, 2017 - 10:18

Bradford Cross, a founding partner at machine learning and big data venture capital fund DCVC, wrote an interesting post on his blog about the impending end to today’s bots hype period, including one section called “Bots Go Bust.” He is right in the sense that hopes for bots will soon crash to Earth, the frothy valuations for AI startups will end, and press coverage will go off the boil.

I want to discuss why the bots-mania took hold, why they were implemented wrongly from the start, and why they will come back strongly — and fairly soon — when the first-wave mistakes are corrected.

The bot bubble

Bot-mania emerged from Silicon Valley and other AI labs as voicebots like Siri, Alexa, and other AI-based products started edging into mainstream use. In parallel, consumers were getting “app fatigue” — too many apps to remember and juggle as brands and startups pushed people to download an app instead of using the mobile web. (Everyone was chasing prime real estate on the smartphone screen.) The bots craze also coincided with the new dominance of messaging, now the preferred way people talk to each other on smartphones.

For bots, the thinking went: Instead of all these apps and interfaces to navigate, people can use their phones more conversationally, not pressing buttons inside apps but just messaging what you want. Bots would work as you message with friends in iMessage, WhatsApp, Facebook Messenger, and so on.

As so often happens, Silicon Valley missed the human element in its excitement about the technological breakthrough. Valley types did not grasp that a bot is essentially a human helper in virtualized form — it is an “agent,” in the parlance of the customer service industry.

When consumers interact with a branded bot, either to buy something or to get service, it is this role as an agent that is the “job” consumers “hired” a bot to help with. And while bots are fast and work 24/7, they are not nearly as smart as human agents. They often fail to understand what you want. They become, in effect, a really low quality human agent.

With all the hype, developers rushed bots into service as standalone experiences. They were thrown into the deep end of customer expectations to sink or swim — and they sank. Most consumers were not very interested in being Guinea pigs for the latest Silicon Valley technology fad (what’s in it for them?). Usage was low, as was customer satisfaction.

As standalone experiences, these bots were not connected to the usual customer care channels and the teams behind them (large companies like banks and telcos have literally tens of thousands of customer support staff answering questions by phone, email, messaging, and so on). That meant, when a bot fails, there is no graceful fallback to human reps. Instead, it’s usually just an “Oops, sorry, I can’t help you” comment.

In this sense, the current wave of bots is a bubble that popped — a moment of excitement followed by a steep drop-off. However, as Cross suggested in his blog post, this will pave the way for the next iteration, which is much more promising. It follows the standard hype cycle of innovative technology: a peak of hype, a trough of despair, then a plateau of useful application.

Get ready to welcome bots back

But there’s reason to think that this plateau is months, rather than years, away. The reason is, first off, that bots are inexpensive compared to human agents, which means that companies have a big incentive to adopt them. And second, there will be a second phase of bots usage bridging the gap between completely automated technology and human support reps.

The error of the sink-or-swim experiment was that it ignored that bots should be specialists, and limited in their roles. When it comes to specific and menial tasks — like updating address or payment cards, or looking up FAQs — they are very good at their jobs, and have no labor costs.

Tango time

What this next wave will do is learn what we call the “tango” — where humans and bots go back and forth within one thread, handing off to each other according to their strengths and weaknesses. Bots will handle simple stuff, like forms, FAQs, and basic processes like requesting an account balance, but they will automatically escalate the thread to a human as soon as a tougher task comes along. If companies execute this dance correctly, the exchange will be seamless, and the consumer will receive a more efficient, intelligent experience.

As bots learn this tango, you’ll see a lot of work handed off to them — around 40 to 50 percent of inquiries from customers at large brands are fairly menial tasks that bots could handle — and bots will go from a curiosity or an annoyance to something genuinely useful.

This post was originally published on

Blog Posts by Category