In 2017, Mezi demonstrated their innovative technology on the Summit stage at The Phocuswright Conference. They took home two awards including the winner of the OAG Award for AI Innovation and the Runner-Up Award for Travel Innovation - Emerging Category.  

Mezi's interview 

 

This article was first published on PhocusWire.com

American Express has acquired artificial intelligence-powered assistant Mezi, which co-founder and CEO Swapnil Shinde says upends the current “DIY” trend of travel search and booking into a “do-it-for-me” model. 

“We use Uber to have someone drive us around, we use Door Dash so someone can get our favorite food and deliver it to us. We believe that the same kind of philosophy will apply to the travel vertical where everyone will have an assistant who will do their search for them, who will always have their back so all they can focus on is their travel,” Swapnil says of the app’s approach compared to those of traditional online travel agencies and metasearch sites.

Founded in 2015, Mezi began as a virtual shopping assistant before pivoting to travel. The app allows travelers to submit requests for flights, hotels and restaurant reservations using text or voice messaging. The app analyzes results from Sabre, Expedia, Priceline, TripAdvisor and others to provide recommendations and instant booking. 

MEZI WINNER IMAGE PHOCUSWRIGHT INNOVATION SUMMIT 2017

Travel concierge 

Shinde says the app suits consumers’ communication preferences too.

“More and more people want to communicate over messaging instead of emails and phone calls, and the same thing will transform the travel industry,” he says. “It’s going to take time, but I think once you discover this mode of communication and this mode of travel research, it’s very hard to go back.”

Under American Express, Mezi – which won the OAG Award for AI Innovation at the 2017 Phocuswright Conference – will become a wholly owned subsidiary of the financial titan.

Through the app, users create a profile with their travel preferences, and the system automatically updates those profiles each time they conduct a search, strengthening the personalization of results.

According to Swapnil, a conversation with Mezi can span a couple minutes, a couple hours or even a couple weeks, “and Mezi will always remember the context of the conversation you’ve been having,” he says. “Every time you talk to Mezi, it knows you slightly better than it did yesterday.” 

To simplify the process for users, Mezi provides just three curated recommendations for each request. Users can ask to see more options or can refine their requirements to generate new results. 

The acquisition follows a $9 million Series A funding of Mezi in 2016 co-led by American Express Ventures, Nexus Venture Partners and Saama Capital. American Express also tested Mezi with a select group of cardholders during 2017.

“Mezi’s AI-powered experience opens up exciting new ways for us to connect with and serve our card members and creates opportunities for us to build more meaningful relationships with them,” said Phil Norman, vice president of American Express Digital Labs.

Bot and human 

Mezi co-founder and chief technology officer Snehal Shinde says one of the platform’s unique features is the automatic collaboration that happens between the bot and human assistants behind the scenes. 

“If the bot is not able to understand something, the bot will hand it over to the human. The human will then train the bot and hand the conversation back to the bot. It’s pretty seamless how this happens and as an end customer you never feel as if you are talking to a bot versus talking to a human,” Snehal says.

Currently about 60% of the requests are fully handled by the artificial intelligence, and Snehal says they want to achieve 80%, at least for business travel, in the next three to six months.

In addition the app interface, Mezi has a desktop experience targeted to travel agents. The Travel Dashboard uses artificial intelligence to help agents process requests more quickly. 

Mezi has logged one million conversations, and the system knows about 50 million words to date. For now it only operates in English, but the co-founders say they hope to add additional languages in the future. 

The winning 2017 demonstration at The Phocuswright Conference

Looking to demonstrate?

Apply to demonstrate your innovation at Phocuswright Europe 2018 or The Phocuswright Conference 2018! 

Information about Phocuswright Europe Battleground and Launch competitions
Opportunities to demonstrate and compete at The Phocuswright Conference 2018