What Gmail's Smart Compose Means for the Future of Work
By Jon Ogden | Senior Manager of Content Marketing at Workfront
“The crucial problem isn’t creating new jobs. The crucial problem is creating new jobs that humans perform better than algorithms.” — Yuval Noah Harari, author of 21 Lessons for the 21st Century
In May 2018, Google announced a new feature for Gmail called Smart Compose. If you’re a Gmail user, you’ve likely seen it. For those who haven’t, here’s a demo from Google’s I/O:
Basically, Google uses an algorithm to analyze the context of your email against their database and predicts the words you might write next, which show in light gray. If you hit Tab, those words turn black, and your cursor moves to the end of the phrase as though you typed it. It’s an expansion on Gmail’s Smart Reply feature, released in 2017, which lists automated responses you can send to people with a click.
In my experience, the suggestions that Gmail gives me are accurate (or at least acceptable) more often than not, and I find that it saves time to just hit Tab and complete a sentence. Since email is something that often gets in the way of the work I want to be doing most, it’s a welcome feature. Mark Cuban, business expert and owner of the Dallas Mavericks, agrees, saying "it adds at least 30 minutes to my day."
Cuban likely isn’t alone. A full 83% of U.S. knowledge workers say that the rise of automation will help us think of work in new and innovative ways, and 69% say automation will give us more time to do our primary job responsibilities.
Chances are that as the capacity to instantly crunch trillions of data points increases, algorithms like Gmail's Smart Compose will become table stakes in a variety of roles.
So, what does Gmail’s latest feature tell us about the future of work?
Here are seven possibilities worth considering so you’re not caught flat-footed.
1. Automated Product Feature Requests
Say you’re a product manager trying to choose a new feature to implement. You’re overwhelmed by your database of user requests, knowing it’s impossible to please everyone. How do you decide what your best next action should be?
Now imagine an automated algorithm that:
- Measures the frequency, sentiment, and passion level of each user request
- Analyzes actual consumer behavior within your software (including everything people search for in the help section)
- Scours your competitor’s websites and third-party review data to see how your feature sets compare
- Maps all that data to key indicators such as stock price, NPS score, and company valuation
- Gives you a suggestion about what to focus on right now, in real-time
Do you hit Tab and take the suggestion? If you do, does that free you up to start inventing completely new features — features that no algorithm could anticipate?
2. Guided Content Management
Every content manager knows that a critical way to drive conversions is to find the right keywords and optimize content for those keywords. So what happens when algorithms are able to read everything written by and about your company while comparing that content with everything written by and about your competitors — all while instantly suggesting topics and keywords so you can hit the number one spot in Google’s rankings for each blog post?
And that’s just the beginning. It may only be a matter of time before Google integrates some form of Smart Compose directly into Google Docs, which opens up the possibility that the more articles you write (especially when using a Google Docs template), the more Google will offer suggestions for each sentence you type, including suggestions on headlines, keyword placement, and relevant sources you might cite to be more authoritative.
If Google doesn’t do it, chances are that companies such as seoClarity, SEMRush, and Conductor will eventually integrate something similar into their product suite since they’ve each already released products in this vein.
All I know is that if Google's algorithm suggested what I should write in this very sentence as I wrote it, I might hit Tab. After all, their algorithm has access to trillions of data points that my brain doesn’t have immediate access to.
3. Predictive Social Media Management
What types of social media posts perform best in your industry? At what times of day? For which audiences? The major social media companies already track this data carefully (as do third parties). Given that these companies have an incentive for users to produce engaging content, why wouldn’t they offer an option to see real-time suggestions in the vein of Smart Compose to guide social media writers once they had the capability to do so — especially if these companies saw that the suggestions amplified engagement?
Of course, such an algorithm would have to adapt quickly, since what’s engaging today isn’t necessarily engaging tomorrow. But that’s the strength of algorithms in the age of machine learning. They always evolve, never sleep. And the more such algorithms were used, the more accurate they would become.
Given that two-thirds of links to popular websites on Twitter come from bot accounts, some may say that in this sense the future is already here. When it comes to social media, we’re already following what algorithms suggest.
4. Personal Assistant for Press Relations
Of course, algorithms aren’t just for digital interactions. They can also help build in-person relationships. For instance, an algorithm in the vein of Smart Compose might look through a press release you compose (possibly with the help of Google Docs) and dig through all the public-facing articles written by current journalists, pairing your exact press release with specific journalists and then spitting out a series of personalized suggestions for connecting in person or by phone to each journalist based on their likes, writing styles, and more.
From there, the algorithm might track which journalists respond and make suggestions about how to deepen the relationship based on what has worked in the past with that particular journalist — essentially serving as a personal assistant with insider knowledge.
5. Personal Assistant for Sales
If a personal assistant with insider knowledge would be useful to someone in PR, it would also certainly be useful for a salesperson. Imagine an algorithm that tracks every bit of online public information about each person you hope to sell to — info such as favorite sports team, favorite music, favorite movies, hobbies, work history, workplace preferences, etc. Now imagine having that information funneled directly into Gmail, where Smart Compose suggests phrases that can appeal to that specific person. You hit Tab, Tab, Tab, and end up with a perfectly personalized email.
The same feature might also help when it comes to leveraging data about the history of a company, so that you see suggested reasons that company should switch from the product they’re currently using to yours (based on how you've trained the algorithm). You’d be able to dash off hundreds of personalized emails a day.
6. Customer Support w/ Machine Learning
The rise of chatbots such as Bold360, which uses machine learning to interact with customers, shows just how deeply automation has already been integrated when it comes to customer support. These chatbots can screen requests before the requests ever need to be sent to a human, dramatically cutting down on resources needed. In addition, customer support software often automatically suggests prompts based on lines proven to de-escalate tension, so that users will say the right thing in the right moment. Finally, companies like Luminoso analyze all text-based feedback about a business (including support tickets, open-ended survey responses, and reviews), and offer guidance about how to best respond.
So where could customer support software go from here? If you’ve worked with upset customers via phone, you know that alleviating tension can be tremendously difficult.
A feature like Smart Compose for such situations might:
- Pull from a database that tracks what customers say as well as the emotion they say it with (via real-time sentiment analytics)
- Map that data to your past approaches (factoring in your emotional state via vocal and biological tracking with an input like a smartwatch)
- Offer up real-time prompts about how to best defuse the exact situation you're facing in this moment.
In this vein, this feature might also note when you need a break to rejuvenate so you can return to the phone in an optimum emotional state. In short, the next frontier in customer support will develop robust emotional analysis to improve the human connection.
7. Coding Integrated with Stack Overflow
Considering how often certain strings of information get repeated when coding software, it’s no surprise that a company like Bayou exists. Bayou takes barebones intentions and then predicts what an engineer would do next in creating a program. “Modern software development is all about APIs,” says Vijay Murali, a research scientist who works on Bayou. “These are system-specific rules, tools, definitions and protocols that allow a piece of code to interact with a specific operating system, database, hardware platform or another software system.”
Murali adds, “There are hundreds of APIs, and navigating them is very difficult for developers. They spend lots of time at question-answer sites like Stack Overflow asking other developers for help.” Software like Bayou reduces time spent researching, so engineers can code. We can be certain that suggestion tools like Bayou will continue to proliferate as algorithms like Smart Compose continue to improve.
Will Humans Become Useless? Is the Future Bright?
Some people might look at these automation services and wonder whether all our jobs will eventually be reduced to hitting Tab. An algorithm suggests something to say to a prospective customer, and we hit Tab. An algorithm suggest our next line of code, and we hit Tab. An algorithm suggests what to say to an angry customer on a chat service, and we hit Tab.
Tab, Tab, Tab.
In addition to potential boredom, one downside to such a world would be that productivity may rise so quickly that businesses wouldn’t need nearly as many employees to do the same amount of work, possibly paving the way for what the writer Yuval Harari calls “the useless class” — a class of people whose skills are no longer needed because algorithms take their place. It’s an idea that governments and businesses should take very seriously, especially since 48% of knowledge workers today know someone who has lost their job due to automation.
However, there’s also a strong case to be made for an optimistic future. After all, Gmail’s Smart Compose feature has freed us up to less time on email. It’s likely that the algorithms of the future will do the same by eliminating all our mundane tasks, just as inventions like the microwave, the laundry machine, and the dishwasher have decreased the amount of time we spend doing housework. The same may be true for today’s algorithmic inventions.
What’s more, seeing suggestions from an algorithm doesn’t preclude anyone from taking those suggestions and doing something even more creative with them. If employees at competing companies see suggestions from the same algorithms, the winners will be those who figure out creative ways to stand out — who know when to hit Tab and when to do something completely different. The best workers will know how to automate mundane tasks and invent something fresh.
As Harari says, “The crucial problem isn’t creating new jobs. The crucial problem is creating new jobs that humans perform better than algorithms.” Those jobs — which will hopefully make full use of humankind’s creative and emotional capacities — will be the jobs of the future. If that’s the case, the future might look bright. It may mean that we’ll each have ample opportunity to explore ideas and theories that have never been thought of before. Leave the automation to the machines, coupled with light human guidance, freeing us up to do whatever work calls to us most strongly.
Given the reality of features such as Gmail’s Smart Compose, it’s the kind of future we should prepare for today.