Languages Magazine

Why Today's Intelligent Assistants Have It All Wrong

By Expectlabs @ExpectLabs

The idea of the omniscient intelligent assistant is already becoming obsolete. Watch our founder identify another approach to intelligent assistants that is much more adept at delivering the information we want, when we want it.


Hi, I’m Tim Tuttle, and I’m the CEO of Expect Labs. As you might know, our company is the creator of a product called MindMeld API, which is a developer platform that lets anyone create an intelligent assistant for their own app, their own website, or their own device. Our company has a very unique philosophy about intelligent assistants. In fact, we think that the companies that are making intelligent assistance today have things backwards.

So, what the big companies are doing today is they’re trying to create this all-knowing intelligent assistant that lives your smartphone; either your iOS phone, your Windows phone, or your Android phone. What most users who use these intelligent assistants know is what you get is an intelligent assistant that’s okay at answering some questions but is generally mediocre at answering most questions. The approach that we’ve taken is very different. What our technology allows companies to do is to create an intelligent assistant that’s very good at answering questions around one domain of content; specifically, inside one application or on one website. Because we’ve taken that approach, what we can generally do for a customer is let them create an intelligent assistant that’s very good at answering questions around a very specific, narrow domain, but knows nothing about other types of questions you might want to ask.

And as it turns out, this is exactly what most users want when they’re using an application. So if a user, for example, goes to the WebMD app on their mobile device, what they really want is very quick answers to questions about medical symptoms and treatment options. They don’t care if the intelligent assistant there knows something about their calendar or their address book. If you go to the Best Buy web site, what you really want is quick answers about what’s the best flat panel TV to buy. You don’t care if the intelligent assistant is also able to recommend restaurants or recipes for you.

That’s exactly what we built our product to do. We made the MindMeld API so that our customers can unlock all of that inherent knowledge that’s in their own content so they can create a really great intelligent assistant that helps their users find the information that they need.

Back to Featured Articles on Logo Paperblog

Paperblog Hot Topics