Clippy gets an AI makeover.
Microsoft Word users from a previous generation are still recovering from the trauma of Clippy, the anthropomorphic paperclip that thrust itself into your writing projects with a cheery, yet unrelenting, insistence. “It looks like you’re writing a letter,” Clippy would ask, again and again. “Would you like help?”
Clippy, introduced in 1996 and eradicated in 2001, was an early attempt by Microsoft to help guide its users, but it backfired because the software couldn’t learn whether you needed help or not. All users were new users to Clippy, even if they were well-familiar with how to use Word.
Users didn’t only object to the clumsy-cutesy design of the feature, but also the attitude behind it. Word processing software was a device for writing, like a typewriter or a No. 2 pencil, not a service. The idea that Microsoft was trying to divine our intent seemed intrusive, almost a violation of the contract between customer and company.
It is with humility, and some amount of courage, that Microsoft is trying again, this time by harnessing the power of machine learning to help Office 365 users manage their schedules.
The product is an addition to MyAnalytics, existing software that allows Office 365 users to keep track of how much time they spent on email, in meetings, and working after hours. In the next few weeks, users will begin to see suggestions about how to better manage their time.
For example, the software will scan a user’s calendar and note that she doesn’t have many two-hour long blocks of unscheduled time, which Microsoft calls “focused work.” In a meeting invitation email, the user might see a line that says “Looks like you have a lot of meetings next week. Would you like to reserve some time for focused work?”
Called nudges—borrowing a behavioral science term popularized by economist Richard Thaler and law professor Cass Sunstein—the tool will also note when you’re sending emails, and remind you if it’s after work hours for a recipient. The language of the reminder—”Consider clarifying the urgency”—was the result of much internal discussion, says Natalie McCullough, who heads the product team for Microsoft. “We don’t want to be too heavy handed,” she says.
Unlike Clippy’s pop-up windows, MyAnalytics’ nudges will learn from users’ behavior, says McCullough, and if users aren’t responding, it will leave them alone. Users can also turn it off.
“We have learned from that experience,” says McCullough, who also notes she wasn’t at Microsoft during the Clippy era. “We’re super careful around the nudges. The advent of AI and ML [artificial intelligence and machine learning] means we can be so much more targeted about when it shows up.”
Much has changed about the world of computing since the life and death of Clippy. We accept corrections to our typing without thinking, depend on autofill, and are now accustomed to—if not entirely comfortable with—programs that read our email and track our search habits. The idea of office software morphing from a tool to a virtual assistant no longer feels far-fetched, and the simple nudges of MyAnalytics don’t seem as outrageous as they might have 20 years ago.
For now, the nudges of MyAnalytics are informed by the email and calendars of the 135 million commercial users of Office 365. But it’s not hard to imagine Microsoft incorporating input from user behavior in Word, Excel, and other Microsoft programs, capturing more aspects of our work lives to create richer, more useful nudges. The end result may be a more productive worker, but also one evermore dependent on Microsoft to function.