4. Meetings!! 4
WSJ - Let's Call Off the Meeting and Get Back To Work (2015)
Andy Kessler - https://www.wsj.com/articles/andy-kessler-lets-call-off-the-meeting-and-get-back-to-work-1420156148
Americans
schedule
approximately
11 million
meetings per day
6. Time
investment Time needed without
Doodle AI bot
Estimated average
Time needed with
Doodle AI bot
Estimated average
Private Dinner – 4-6 Participants 20-25 minutes 5-10 minutes
Business Meeting – 6-8 Participants 26-30 minutes 13-17 minutes
Private Dinner – 10-15 Participants 90-120 minutes 31-35 minutes
96 people were selected to use Doodle and alternative scheduling tools
https://en.blog.doodle.com/2013/07/08/study-reveals-time-spent-with-scheduling/
7. Iteration
Iterations
Estimated average
Time needed
Estimated average
Private Dinner – 4-6 Participants 8-12 14-17 minutes
Business Meeting – 6-8 Participants 12-16 20-25 minutes
Private Dinner – 10-15 Participants 10-15 46-50 minutes
96 people were selected to use Doodle and alternative scheduling tools
https://en.blog.doodle.com/2013/07/08/study-reveals-time-spent-with-scheduling/
8. Calculate
Up to 50% of these
meetings are a waste
of time, costing the US
$37 billion per year
8
https://hbr.org/2016/01/estimate-the-cost-of-a-meeting-with-this-calculator
WSJ - Let's Call Off the Meeting ..Andy Kessler - https://www.wsj.com/articles/andy-kessler-lets-call-off-the-meeting-and-get-back-to-work-1420156148
11. Challenges in Bot
• Business users have very little tolerance for mistakes
• Good AI needs a lot of data, maximum organisation
don’t have it
• Scheduling scenarios are complex, like one to one, time
zones, venue availabilities etc.
• Who will cover the cost?
https://hbr.org/2017/07/how-we-built-a-virtual-scheduling-assistant-at-microsoft
12.
13. A ScheduleBot – Microsoft way
• Start with lo-fi paper prototypes to a group of users to
rapidly collect user feedback
• Next build a wire-frame mock-ups to test them with
users in a slightly more realistic setting
• Wizard of Oz, where UI is real but backend is human
researchers
https://hbr.org/2017/07/how-we-built-a-virtual-scheduling-assistant-at-microsoft
14. A ScheduleBot – Microsoft way
• Selected random users from various companies to signup
• Added Virtual Assistant in email Cc on meeting invite
• AI suppose to look for schedules and create calendar invites
• In reality employees were doing calendar booking
• It enable Microsoft to get a product into people’s hands early
• Learn deeper understanding of the problem schedulebot may
face
https://hbr.org/2017/07/how-we-built-a-virtual-scheduling-assistant-at-microsoft
15. Microsoft’s Calendar.help
• Learning users expectation from schedulebot
• Help to design workflow with microtasks
• Continue humans-in-the-loop model and hired a staff
of workers to perform microtasks
– One suggest optimal times, another looking at available
locations, and another working to reschedule the meeting
if needed
https://hbr.org/2017/07/how-we-built-a-virtual-scheduling-assistant-at-microsoft
16. Microtasks helped to focused worker’s attention on
one thing at a time01
02
03
It helped to collect high-quality data which was used
to automate the scheduling processes
It helped to create variety of machine learning models
to understand natural language
17. Other ScheduleBots
•X.ai (Amy): $39/month (all prices in USD)
•ClaraLabs.com (Clara): $499/month
•Mimetic.ai (Evie): $25/month (Doesn’t offer custom-domain)