No, he wants to show me angel.ai. Hadzaad's fifteen-person New York City company is tackling two projects right now: a refocused GoButler, and the launch of angel.ai, a business service focused on AI automation - or, more accurately, NLP-powered services. So we sat down, and Hadzaad didn't hold back.
The next phase starts with GoButler's fully-automated flight search and discovery service. That's right - no more "anything goes" virtual assistant. GoButler is now a completely automated flight discovery service - and Hadzaad wants me to try it. So I type in a complex itinerary that has me heading to Bangalore and back in a matter of a few days:
Navid Hadzaad: You tried to break it.
Jon Reed: That's my job, dude.
Hadzaad: See - it extracts the flight parameters you wanted. We are working with different airlines now. I can't announce which ones, but does that make sense to you?
Reed: Yes - travel bookings are a constant pain.
Hadzaad asks me to type in a more open-ended query, so I choose a round trip from New York City to Bangalore; itinerary options are cranked out in the chat window. The automated assistant responds: "Ok, when do you want to return? Or let me know if this is a one-way trip." (Note: GoButler's flight service is in "apply mode," but you can demo this flight functionality on angel.ai). The tech behind Angel's flight service is the foundation for other NLP business services.
How did we get from virtual assistant to an NLP service?
So how did we get from a manual virtual assistant to a fully automated service? Hadzaad:
For me, the goal with GoButler was to figure out how people are using these type of services that have a conversational element to them... One of the big reasons for starting GoButler was to aggregate data, and then to figure out which categories you can start with.
Why not a hybrid model?
I don't believe in a hybrid model. Hybrid models don't really work. They don't work from the agent perspective, and they also are horrible from a user perspective.
That's two big "pivots," if we can stomach the term: manual to automated, and consumer (GoButler) to business (Angel)
So if angel.ai is the new focus, what does this mean for GoButler? "Let's see - it's tough for me to say," concedes Hadzaad.
Chatbots should be solving problems, not "creating experiences"
Hadzaad's is bullish on Facebook Messenger - but highly critical of their chatbots:
We've been working this for two years... It's annoying when you see a lot of new people coming in with no background, getting excited about chatbots... If you look at the chatbots that were launched with Messenger, would you ever use them?
For Hadzaad, chatbots shouldn't be about "a great interactive experience." They should be about getting a task done:
Did you see the 1-800-Flowers bot? They didn't think about, 'Okay, how do we create a super efficient flower ordering experience in Messenger?" They just thought, "Okay, let's create a chatbot for ordering flowers," which doesn't make any sense. It's just tedious.
It's about solving an issue quickly:
I don't see a lot of value in chatbots. I do see a lot of value in using natural language commands, like we just did with flight discovery, to get you where you want to be. You don't have to do a lot of work. I don't see the value of having a back and forth.
1-800-Flowers didn't start by solving the right problem. The right problem is, "How do I order my mother-in-law flowers in thirty seconds?"
Enter messaging - "the new operating system"
Hadzaad isn't obsessed with Facebook Messenger in particular; he points to WhatsApp in Europe, and WeChat in China, which is much further ahead with embedded chat services. He's drawn to the dominant chat platforms; they have the data and user experience locked in:
I am a big believer in Messenger becoming the new operating system... They have all my data because it's in Facebook; I don't want to download another app. That's the biggest problem they can solve, and I believe that someone will, whether it's Messenger or whether it's iOS or Google.
Enter messenger-based applications:
I don't think chatbots are going to play a big role there. I think people want the most efficient experience. A lot of times, the most efficient experience is going to look very native... For example, any Messenger conversation that I'm in, there's a transportation button, right? They will essentially open the Uber or Lyft app for me. That's all it is... It has nothing to do with the chatbot or anything like that.
"None of the airlines or hotel companies are near to developing anything like this"
So what is angel.ai's best case scenario in two years?
We power the technology for a lot of different business on messenger interfaces... We want to solve an interface problem on mobile, and we want to do that with natural language. We're actually bringing some brands on Facebook right now with angel.ai.
He wants to enable companies to jumpstart their NLP services:
I see a lot of potential to bring brands on existing big platforms like Messenger using our technology. None of the airlines or hotel companies are near to developing anything like this.
Fine - but what sets GoButler apart?
We have millions and millions of data points about people having commercial intents in a messaging interface. Then we have our own secret sauce. We're data scientists, and we're working with the data at the human level. We've done this for flights. We've done this for other domains. This is super valuable because no one has this kind of data, and no one has this particular technology on commercial-focused messaging/chats. That's hard stuff. You can't do that overnight.
He hopes the flight discovery demo is a stark contrast to chatbot "fluff":
We have the demo out there is because there is so much fluff. Even people who are not techies, if they try this out, they say, "Oh, my God. This actually understands me." There is nothing comparable out there.
Hadzaad is adamant: he was always trying to build a technology company. He's rejects the "GoButler the Magic Clone" narrative ("I never wanted to build a concierge service"). Humans end up bogging down e-commerce transactions - nor is he a fan of messenger-based customer service. Hadzaad believes it's better to keep the escalation/exception channel separate ("If you want to book twenty people on United Airlines, you still need to call them").
I tried to get Hadzaad to wax nostalgic about his ex-Rocket partners not involved in this phase, but he didn't bite ("Thing change... but there's no bad blood"). Hadzaad warns against trying to merge friendship and startups anyhow ("Don't make things emotional. Take money from people who know about the risk.")
Which brings us to his investors. Are they weary of all the course corrections? On the contrary, asserts Hadzaad: "They're excited to be a part of this space." And that's where his prior startup experience comes in:
I've done other startups before this. I've done very successful ones. I've also done really shitty ones. The most important thing is: leverage the flexibility you have. Have conviction, but also don't be afraid of admitting that something's not working.
If you want to figure it out, don't have too much pride. Learn. Learn quickly and adapt, that's important for any startup. You can make a big mistake by being too attached to something.
And with that, Hadzaad rushes on to the next, leaving me to question my own views about app design. For GoButler and angel.ai, it's still early days, with no promises made. In other words, a story to watch.