No worries! That’s the bot talking, offering a breezy response to a mildly apologetic email: Your coworker wants to reschedule a meeting? Sure thing! And they’ve proposed a new time? That works! If you’ve opted in to Gmail’s Smart Replies, these exchanges should look familiar. The AI-generated shortcuts at the bottom of an email promise effortless efficiency in exchange for a tiny piece of your digital humanity. But us humans are proving eager to make the trade: More than 10 percent of all replies on Gmail now start with a suggested Smart Reply. Cool, thanks!
The apps we rely on to stay productive at the office are being infused with ever larger helpings of artificial intelligence. They’ve been getting smarter for awhile, but recent advances in cloud computing, neural networks, and deep learning have sped things up. “When we first did spam filtering in Gmail 14 years ago, we were using algorithms that were sophisticated at that time. Our AI techniques have changed a lot since then,” says Rajen Sheth, director of product management for Google Cloud AI. “Things like Smart Reply and Smart Compose have really brought this to the forefront.”
Google’s Smart Compose is the show-off cousin of Smart Reply. It actually tries to complete phrases for you as you’re typing an email, and it’s uncannily good at predicting what you intend to say next. (“Thanks,” I started to write to a Google spokesperson. “This is really…” “helpful,” Smart Compose suggested.) Other products in G Suite, the company’s assortment of work-related apps, are peppered with assistive features, too. In Docs, the Explore tab uses machine learning to suggest documents, emails, or web links related to the topic you’re currently working on. The mobile version of Calendar lets you create goals, then automatically jams them into your schedule for you. Thanks to AI, you’re no longer sorry-too-swamped to file your travel expenses on time.
Microsoft, eager to shed its image as the stalwart provider of ho-hum desktop software, has been imbuing Office 365 with AI-powered features. Word now knows when you’re making a to-do list and tracks those items as … to-do items. PowerPoint uses computer vision and machine learning to color-match your slides with the hues in imported photos. And thanks to image recognition tech, you can snap a photo of a data table, import it into Excel, and end up with a fully editable table. Clippy was never so skilled.
Microsoft has suggested that the future of AI in the workplace may be one in which your email app can sense, by your tone, that you’re frustrated. But that doesn’t mean the company is hurrying to suggest email etiquette, or plans to ask if you’re really sure you want to hit Send. That’s partly because Microsoft says it wants people to remain the heroes of their work domains, for now. “The bigger part of our responsibility is figuring out how to keep the assistant in a more humble position, providing just a whisper of a recommendation,” says Ronette Lawrence, Microsoft’s AI planning and user research lead.
Tech behemoths aren’t the only ones automating onerous office tasks. Cloud-first newcomers have been helping to establish what work looks like in the age of AI. Bots have become a near-essential part of Slack’s collaborative chat software; every so often a Slackbot will suggest you leave a channel you haven’t visited in awhile. The Seattle-based startup Textio is wielding machine learning to help recruiters and HR managers write more compelling job postings. Thanks to AI, the company has learned that using “AI” in an engineering job post is less effective than it used to be. “Big data”? So five years ago.
Perhaps because of this, Textio cofounder and chief executive Kieran Snyder, a natural language processing expert and former Microsoft product manager, prefers to call Textio an “augmented writing” platform. “There’s been a lot of stuff that’s aimed at making your work look pretty, or collaborate easily, but nothing that made words work better,” she says. Snyder claims the average time to fill an open job role is about 10 days shorter for clients who use “growth mindset” phrases suggested by Textio’s software.
Textio’s larger implications are staring users right in the face, no bot guidance needed: The tech doesn’t just speed up the hiring process, but also diversifies it. The software recognizes phrases that tend to attract more women job applicants, and alternately, more men. For one Textio client, the phrase “work independently” has driven a 27 percent higher rate of job applicants who identify as women. Phrases like “ninja” or “rock star” skew more towards male applicants.
The question with AI-driven work tools, then, particularly ones engineered to identify potential biases, is whether they help solve for biases or perpetuate them. The answer is likely both. Plenty of computer scientists have determined that biases in AI algorithms are inevitable; machine learning models will simply adopt their human authors’ own worldviews. Relying on a bot to handle your messaging might seem like it would produce a neutral outcome, but like you the bot has its own baggage. Late last year, Google opted to remove gender-based pronouns from Smart Compose after a company researcher discovered, during a beta test, an email about an investor meeting defaulted to the pronoun “him.” (Google says the live version of Smart Compose never exhibited this bias.)
Beyond that, AI-powered work apps raise questions around privacy (you knew your boss had system admin access to your work, but what does your bot know about you?) and the small matter of humanity. What do chats and emails, art forms in their earliest days, look like when we’re all exchanging the same rote AI-penned messages? Will our resumes be indistinguishable from one another? Will our slide decks all look the same? (OK, they already do.) Will WIRED articles drafted in AI-powered writing software all read the same? Microsoft’s Lawrence says she believes the opposite could prove true—that the more boring tasks AI can handle for us, the more free time we humans have to be creative. Sounds good, thanks.
Lauren Goode (@laurengoode) is a senior writer at WIRED covering consumer tech.
This article appears in the March issue. Subscribe now.