Independent Thinking Blog

Here’s Why Chatbots Aren’t Ready for Prime Time

It’s time to talk about chatbots again.

Last year I asked if chatbots would solve customer service.

As companies, we want automation to take over simple tasks. But, as consumers, we want something else. We want seamless transactions and brand consistency.

We can already book a flight, make dinner reservations, purchase clothes, check our bank accounts, and do a lot more without talking to a human. Mostly we do this via websites and mobile apps. The chatbots are supposed to add an AI layer. But for what purpose?

And (and this is a big “and”), do consumers even need them?

Why build a chatbot?

I’ve tested out a couple of news-focused chatbots.

Al Jazeera English has a Facebook Messenger chatbot that sends you a daily briefing. You select your regions and topics, and voila — headlines with links to stories. The problem is that this is really just a glorified e-mail newsletter in SMS format. While it’s very important for brands to experiment with SMS as a communications channel, it’s not an AI function.

Quartz is a standalone news app that operates like a messenger app. It delivers bite-size news, and gives you an opportunity to get a second bite-size paragraph or two. A little more chatbot-like, but still not particularly innovative.

In other words, what’s the point?

We’re at chatbots 101.

chatbots 101

This isn’t what we need.

A recent article on chatbots argues that they’re just not that smart yet:

“AI, despite all that it can do, is still in its relative infancy, and a company has to be careful not to ask it to perform beyond its current abilities.  Train chatbots to handle simpler tasks and processes… those that are repeatable, structured and unambiguous. Monitor call centers and study the transcripts of “live” agents to identify those simpler tasks and processes. And if a task proves to be too complex, hand it off to a human — with the bot serving as a partner by providing support where it can.”

Companies want chatbots to solve customer service, or be virtual assistants, or do more complex if-this-then-that tasks, but they aren’t programmed for these things.


The chatbots are failing. For now.

Chatbots: 1st attempt at learning

Facebook has scaled back its use of chatbots following a 70-percent failure rate. We don’t know why we need them. Plus they’re still pretty sterile, divorced from the rest of the brand experience. There’s little differential value in them right now — and they’re not fun to interact with.

Amazon’s Alexa is more fun. (Actually, I quite like her.)

That said, don’t give up on chatbots.

Mitch Joel writes:

“Brands should not give up. We have seen data and reports like this before. It’s classic for any nascent technology. It looks like people are having a bad experience, so it’s painted with a dead-on-arrival headline… Chatbots may be struggling with the technology and the adoption today, but they will become mainstream tomorrow.”

Watch this space. I’m not sure how we’re going to end up deploying this technology, but businesses will figure this out.

101 by Steve Bowbrick (Flickr); Fail by MrsDKrebs (Flickr).

You May Also Like