For Complex Queries, Chat Will Replace Search Engines

Chat interfaces have critical advantages over a traditional search box’s “fire and forget” interface

Chatbots had a brief and glorious moment earlier this year.

For a few weeks, particularly surrounding Facebook’s announcement of its new Messenger API, chatbots were on the tip of every Medium-post-writer’s tongue. Chatbots for shopping! Chatbots for weather! Chatbots for everything you can do on the internet, and more!

And then the hype…subsided. The journalists had decided: chatbots just don’t work. It felt like one of the briefest mini-bubbles in tech history.

Was it a Dutch-tulip-style bubble, though, or a dot-com-style one? In other words, was the excitement purely ephemeral, or a sign of things to come?

Based on the title of this post, you know my vote. Chat interfaces are immediately comprehensible because talking/writing/texting is how we “execute” any task that involves other humans. Automating a text conversation is a obvious next step in human-computer interaction, because humans intuitively “get it,” and computers are getting better and better at figuring it out.

But what kind of bots are best? Where do they show significant advantages over the tools we have today? I have a few ideas, but today would like to focus on one: complex searches.

Where Current Search Engines Fall Short

Search engines are a de facto interface to all kinds of information today: restaurant information (Yelp), music (iTunes, Spotify), job opportunities (LinkedIn, Indeed), and even the entire internet (starts with a G). They’re a great way to get exactly what you need, as quickly as possible.

But sometimes your needs are too complex to be reduced to a simple query. That’s where the opportunity lies—in areas where search engines aren’t up to the task.

Complicated Needs

Say you’re looking for a job, and want to provide more information than just the job title and city. You’re out of luck on Indeed.com.

Or you want something to eat, but you have one vegetarian in the party, and your friend is coming along who just broke up with his girlfriend, so he wants a stiff drink (and another one at a cocktail bar after). Have fun finding that on Yelp.

Or you want a certain item of clothing, and have no way to describe it other than that it’s similar to another item you wear. Can you find that on Google?

Vague notions, complicated constraints, or ineffable traits are nearly impossible to communicate to a search box. If your needs can’t be distilled into a simple phrase or filtered by a few checkboxes, a search engine can’t help you.

But a messaging-style interface is perfect for these applications. The agent on the other end of the line (human, bot, or hybrid), can ask questions until they know exactly what you want, and then go find it for you. And if they don’t solve your problem on the first try…

Feedback

They can learn from you, and try again! While Google improves (gradually) based on what you click on, most search-engine interfaces don’t—every search on Kayak or Yelp is a fresh start. Even the best machine-learning algorithm based on just your clicks can’t improve as quickly as a piece of software that explicitly collects nuanced feedback and rapidly adjusts. The feedback is simply richer data.

Conversational search tools will be wholly oriented around learning and improvement based on feedback, and because we humans are used to constantly giving and receiving feedback “in real life,” we’ll happily comply.

Periodic Updates

A conversational search tool can send periodic updates when new results match your query. Using the job example, if new job opportunities come online, you don’t have to go back and search for them again—the bot will send them to you periodically.

Some search tools do this today via email (e.g. Hipmunk), but chat is more information-dense, easier to sort through, easier to interact with, and allows for instantaneous feedback (see the point above). In many ways, chat interfaces do what people already do in email today. They just do it better.

Empathy

This is a bit of a touchy-feely one. It’s hard to explain, but, even if you’re talking to a search tool that is clearly a bot, the experience is a little bit more human. Using language, rather than keywords, makes the interaction feel more compassionate, more empathetic. As natural language processing/ response tools get better, this will become even more apparent.

While Google is often able to help you out, it never feels like it’s “on your side.” Chat-style search interfaces of the future will.