July 1, 2013
Dennis buys the x.ai URL at an auction. Expensive. Ouch.
Conversations with Alex about killing the inbox lead to the conclusion that meeting scheduling accounts for much of the bloat. Dennis actually counts how many meetings he scheduled in 2012. That number, 1,019, represents about 8,000 emails, and that’s not counting reschedules.
Dennis brings the band back together to hack away at a machine-driven solution for meeting scheduling. Mostly, Dennis, Alex, Matt and friends are fooling around on the whiteboard, trying to understand whether the meeting-scheduling universe is a bounded space. Dennis also discovers that the purchase of a single yogurt buys four undisturbed hours at the Union Square Pret a Manger.
December 9, 2013
Dennis and Alex start scheduling meetings for each other. Each feels a great sense of relief when he gets to send a meeting to the other one. They are trying to see if this could be done without context. They also want to find out if the data clusters in meaningful ways. They both operate under the handle firstname.lastname@example.org.
February 3, 2014
Dennis and Alex restructure the experiment. Now Barbara schedules their meetings with no context save what she receives in the scheduling conversations. Barbara takes over the email@example.com handle.
A week later, several dozen emails go out to the first alpha testers. Now Alex and Dennis begin to focus on the team they would need to build to get an AI meeting scheduling company off the ground.
Over the next two months, the experiment is expanded. In all, 45 friends accept the offer to use Amy. Two human assistants do the work of scheduling while Dennis and friends analyze these scheduling-related conversations.
April 14, 2014
x.ai is incorporated. Founders Dennis R. Mortensen (CEO), Alex Poon (COO), Matt Casey (CTO) and Marcos Jimenez Belenguer (Chief Data Scientist) move from Pret a Manger to a few desks at the LiveIntent offices on Church Street.
Seed round ($2.1M) closed. Funds dedicated to answering the question: Is meeting-scheduling a tractable problem? The team is going to prove that a machine can learn to schedule meetings or pack up and go home.
Begin the work of creating a definitive model of the meeting scheduling universe. It’s way harder than the founders thought it would be.
First hire: Laney Caldwell, AI Trainer. She’s still here (and now a Product Manager).
Method to capture data and simple annotation console built, as well as simple response templates. The first beta users are admitted. Now, scheduling is an annotation task—data in incoming scheduling emails is tagged so that the machine can begin to learn.
The answer is yes. Meeting scheduling is a tractable problem.
Series A raised ( $9.1M). Use funds to build up data science and engineering teams.
Still working on defining the meeting scheduling universe.
Andrew, Amy’s brother, is born.
Amy gets asked out on a date.
Series B financing raised ($23M).
Have more or less defined the meeting scheduling universe, 2.5 years after the team began. Full set of annotation guidelines complete, after countless revisions. These run over 120 pages.
Hiring data scientists, engineers and business leads at a rapid rate.
… to be continued