There’s a big debate going on in policy and tech circles over whether or not artificial intelligence will turn out to be a good thing or a bad thing for society.
One side sees AI primarily as a job destroyer, while the other sees it as a means to liberate workers from mindless administrative tasks. Both sides are missing a bigger point: AI is coming one way or the other; what matters is how we respond.
The onset of AI in the workplace raises instead a new set of far more important questions that deal more directly with this reality: How can we use artificial intelligence to help us constantly get better at our jobs, learning necessary new skills along the way? How can AI be used to help workers rise above the mundane tasks it is automating away?
The answer is something we’ve dubbed the “coaching cloud,” and it forms the foundation of a major advance in how I think businesses will use software to augment the capacity for human learning. I also believe it will drive the creation of the next generation of iconic enterprise software companies.
For 40 years, business software has essentially replaced processes that previously required paper forms. I worked with Marc Benioff in the early days of Salesforce.com, and though delivering multi-tenant software over the web was a huge breakthrough, for users, it was largely the same experience of forms. I think dynamic coaching will finally replace the old forms-based paradigm. Instead of giving us static workflows and reporting tools, this new software will help us get better at our jobs.
As AI capabilities improve, we can either treat it as a crutch that relieves us from thinking — examples include Waze and Google Maps — or as an asset that helps us use our brains more effectively and creatively.
My hope is that we choose to win the race against automation by choosing the latter. Otherwise we’ll be overtaken by the automation curve: Human workers will simply be cogs in business processes driven by machines. At that point the race will be lost. It may sound extreme, but the coaching cloud gives humanity a fighting chance.
A coaching cloud company uses machine learning to guide workers toward doing their jobs more effectively while they’re doing it. The key ingredient of the coaching cloud is software that gathers data from a distributed network of workers, and identifies the best techniques for getting things done.
The software acts as a real-time, on-the-job coach, guiding employees to successful outcomes, and in the process gathering new data that’s then fed back into the system. Rather than dispensing “one-size fits-all” advice, it instead offers coaching that’s uniquely tailored to each worker and the task they’re doing at any given moment.
Coaching cloud software gets better over time by learning the best practices that are proven effective across a variety of situations, identifying those outlier cases where a creative person finds a new, better solution, and adds those techniques to its coaching. This allows others to learn from the experience of those more creative workers. This is how humans become the “mutation engine” in this evolving process, generating new ideas which in turn benefit everyone else.
The coaching cloud can also help solve other problems: Remote workers can be trained more easily. Where you work will become less important than how good you are at your job, and how flexible you are at improving your skills.
We think coaching cloud companies will rise in nearly every sector of the economy, especially related to sales and service jobs where human-to-human interaction is essential. Interpersonal communication is dynamic and nuanced in ways that only humans can fully grasp, and it changes from place to place. It can never be fully automated, but machines can coach humans in ways that improve our decisions.
There are already some companies that are showing the way forward, and I think they exhibit some core principles that will set successful coaching cloud companies apart from the pack.
Go deep, not big, with data
Successful coaching cloud companies focus on a narrow set of problems within a specific domain or field; they avoid the broad “apply AI to everything” approach popular among many startups today.
In order to succeed, these companies will need access to huge troves of user-behavior data specific to the problem they’re trying to solve. That means they’ll need a network to generate that data, one that’s highly focused on the specific jobs at hand. In this case “deep” data trumps “big” data.
Textio is a great example of a company that does this well. It uses machine learning techniques to help businesses write better job postings that are more likely to attract qualified candidates. Using data created from the analysis of more than 150 million job posts and outcomes, Textio’s system can accurately predict how likely a post is to get the attention of the right applicants. And it’s making a difference in how companies hire. One customer, Expedia, saw a 25 percent increase in qualified applicants coming through the door, and a 20 percent increase in women applying for technical and management jobs.
It works alongside the person writing the post, helping them craft the most effective post they can, and learns from the unseen words and phrases of outlier writers. Textio calls this “augmented writing,” and it’s easy to see how the same capabilities may be applied in other areas of business writing. Critically, however, Textio has started with a narrow focus on job postings, allowing it to amass the world’s largest data set in this domain, and thus, the best coaching.
Sweat your data, not your algorithm
The best coaching cloud companies will use open source tools like Google’s TensorFlow, Microsoft’s Cognitive Toolkit and Amazon Machine Learning. Don’t obsess over creating a unique algorithm — that isn’t your company’s defensible intellectual property: Your proprietary user behavior data is.
Chorus has created an intelligent engine that listens in on every call that a sales rep conducts with a prospect. During these calls, it offers coaching on what phrases and words to say and when to say them. The aim is nothing less than raising the odds of successfully closing a deal.
The system may suggest asking certain questions or a change in tone of voice depending on the type of customer on the call, or even flag that you forgot to answer a customer’s earlier question. It then looks at the outcome of the customer interaction — closed deal or not. Relevant data is added to the corpus of knowledge that forms the coaching network, which in turn helps guide other sales reps.
What Chorus is building is very difficult to do well, but it is ultimately the quality and quantity of proprietary data it gathers which determines the quality of the coaching. The early results are encouraging. One customer, Everstring, saw a nearly immediate 4 percent increase in its rate of closings.
Make it useful, make it visible
Successful coaching cloud companies must work hard to build a user experience that encourages its use. Coaching UIs that are intrusive or annoying discourage use. Imagine if your Fitbit vibrated with every possible alert; you’d throw it in the trash. Lack of use leads to a lack of fresh data, which is deadly for this new networked software.
Also important: Build a UI that lets users know that what they’re doing is helping their colleagues; this encourages further contribution. It’s one of the reasons that Waze works.
Guru has created a clever Chrome browser extension that links workers to the institutional knowledge they need to complete certain tasks. Inside every company there are tasks that require a unique workflow. It may be how to handle a product return or how to address certain customer objections in a sales process.
This knowledge tends to get scattered into any one of several miscellaneous documents on a corporate intranet or wiki site, but it mostly lives in the heads of employees. When Guru notices someone doing one of these tasks in Gmail, Salesforce, Zendesk, Slack or other applications, it automatically surfaces related information, in context, and in real time.
Employees — especially those who are new on a job — like it because it saves them the time it takes to look up the information they might need, so they keep using it. The high rate of usage creates more valuable data on what works best, which helps Guru make better suggestions over time. Since deploying Guru, Shopify has seen a five times increase in knowledge-base usage, speeding up critical processes. Intercom has seen a 60 percent reduction in the time it takes its support team to respond to customers.
As the coaching cloud takes hold in large enterprises, we may finally see an enterprise software company ride the network effect with its data to achieve significant scale.
The big gorilla in the space is Salesforce.com, whose market capitalization recently passed the $70 billion mark. As successful as it is, Salesforce is much smaller than Facebook ($500 billion market cap) and Google/Alphabet ($650 billion).
There’s a reason for this disparity: No enterprise company — with the arguable exception of LinkedIn — has yet fully harnessed this network effect with their data the way the consumer internet giants have. I think the first enterprise company to do it will come from the coaching cloud, and will dwarf today’s gorillas.
Gordon Ritter is a founder and general partner of Emergence Capital Partners. Prior to founding Emergence, he spent more than 15 years founding and building companies that pioneered new markets, including embedded web-based interfaces, server appliances, “on demand” services for the SMB market and web-native application development. In 2003, Ritter led Emergence Capital’s first investment, in Salesforce.com. Reach him @gordonritter.
This article originally appeared on Recode.net.