The Problem with Chatbots

The chatbot countermovement is well underway, leaving many devotees to wonder what happened to the promise of tireless brand ambassadors who could supercharge the sales funnel and ease the load of beleaguered customer service teams.

While reality has so far fallen short of the vision, there are plenty of reasons why true believers should hold off unfriending their favorite "Hello World" Messenger bots just a little longer. Below are just a few arguments I hope serve you well in your attempts to defend chatbots from all the haters on the relevant corporate Slack channels.

1) Chatbots have us right where they want us... on the Hype Cycle. This isn't the first time we've been through this (see: blockchain, AR/VR, 3D printing, etc.) but somehow we always over-exaggerate the effects of the Trough of Disillusionment. The key question isn't whether chatbots are dead. The key question is what other technologies could force chatbots to power down, and the answer to that is simple: there aren't any yet. So chin up chatbot fans, lacking a competing technology chatbots ultimately can't lose as long as they provide more value than the alternatives, which at the current moment consist exclusively of expensive, slow, and error-prone human-powered solutions.

2) Chatbots are just a channel. As anyone who has spent 15 minutes on Chatfuel can tell you, creating the interface is easy. What's really holding chatbots back is the state of AI, whose job it is to interpret what the bot should do with all that information it's collecting. In other words, chatbots are really just a leaky abstraction of AI.

It's no knock on the AI community to say that AI is still largely unprepared for prime time considering only a few years ago the whole topic was better suited to Analog Magazine than Hacker News. AI will catch up, but in the meantime to reduce our collective disillusionment I suggest re-evaluating where chatbots work best. It may be painful to accept that broad, open-ended journeys are just not suited well to chatbots just yet. But chatbots are fantastic when paired with well-guided journeys through largely-static option trees. Consider how many Facebook Messenger bots lean heavily on buttons and prompts versus free-form text entry. Why is this so? Because as much as marketers want omniscient chatbot agents that can handle any input seamlessly, users just want to get to the end state as efficiently as possible. So for now we're stuck with the "Why don't you just tell me what you want?" type chatbots. And that's fine for now.

3) Anyway, why are you stressing? Chat has already won. We chat on our phones via SMS and Messenger. We chat at work via Slack and gchat. I've been writing chat applications since at least 2003, but most of them wouldn't look like a chat application because the bot itself is hidden away from the end users. In that way chat "winning" isn't about convincing people to adopt a new technology. It's about exposing the existing but largely hidden technologies to end users directly. And that we'll only get better at with practice.

So chat on true believers. We may yet have dark (quiet?) days ahead of us, but ultimately chatbots will become so pervasive that we'll wonder how we did anything without them.