How Companies Are Using Machine Learning to Get Faster and More Efficient

Machine-reengineering is a way to automate business processes using machine learning. Although machine-reengineering is new, companies are already seeing striking results with it, particularly in boosts to speed and efficiency. Studying 168 early adopters, we’ve seen speed improvements of two times or more for most business processes — and some organizations are reporting speed improvements of 10 times or more.

How do companies do it? Our study found that organizations are using machine-reengineering to establish new forms of human-machine collaboration that break through the bottlenecks of complex digital processes. In some cases, such as interpreting images or writing reports, machine-reengineering directly helps workers perform digital tasks. In other cases, machine-reengineering helps people uncover insights that are buried in a mountain of data. Here are some examples of how companies are using the speed and smarts enabled by machine-reengineered processes.

Scanning Images, Voice and Text

As companies implement digital strategies, they introduce new labor-intensive tasks to sort through all the data they’re collecting. This data is highly unstructured and produced in a variety of formats at an ever-larger scale, which requires people to arduously scan through it for specific items to complete a single process step. Human-machine collaboration focused on digital-data scanning can accelerate at least three kinds of routine digital tasks.

Previewing video. Clarifai, based in New York City, uses machine learning to find people, objects, or scenes in videos in far less time than a person can. In one demonstration, a 3.5-minute clip was analyzed in just 10 seconds. The technology can pick out kinds of people — mountain climbers, for instance — to help advertisers more efficiently match ads to the videos. It can also be used to help editors and curators find new ways to organize video collections and edit footage. This kind of auto–editing assistant can dramatically change the day-to-day tasks of workers in media, advertising, and film.

Interpreting images. MetaMind, in Silicon Valley, offers a service called HealthMind, which uses computer vision to analyze medical scans of brains, eyes, and lungs to find tumors or lesions. HealthMind relies on deep-learning techniques for natural language processing, computer vision, and database prediction algorithms. The upshot of HealthMind is that doctors spend less time interpreting images and more time consulting with their patients.

Documenting and data entry. Machines can learn to perform time-intensive documentation and data entry tasks, letting knowledge workers spend more time on higher-value problem-solving tasks. The London-based startup Arria, for instance, helps its customers automatically generate reports in industries ranging from health care to finance to oil and gas. The company’s natural language processing technology learns how to write reports by scanning texts and determining relationships between concepts. Then it scours incoming data to build new reports. Arria has found that the process changes can increase knowledge workers’ productivity by 25%. Engineers, for example, have saved up to 40 hours of reporting task time each month.

Uncovering Buried Insights

Increasing the amount of data in a workflow can increase the amount of time needed for insight and action. We’ve seen this in stock trading, marketing, and manufacturing, where more data streams make it harder to find information that is urgent or meaningful. With machines as sidekicks, though, people can more quickly find valuable insights buried in big data. Our research found companies demonstrate this in at least four types of analytical tasks.

Market monitoring. Dataminr, based in New York City, uses a variety of indicators to identify tweets containing relevant information for stock traders. By monitoring the   propagation of information throughout the network, Dataminr evaluates relevance and urgency. An alert sent to a trader that provides even a three-minute advantage can translate into significant profit. News services are using Dataminr to find breaking news, which lets reporters cover stories faster.

Predictive modeling. SailThru, also out of New York City, helps marketers deploy more effective promotional emails by analyzing email and web data to build customer profiles. SailThru’s system learns customers’ interests (biking versus rock climbing, for instance) and purchasing behaviors, and then predicts which individuals will make which purchases and when, delivering the right message when it’s most effective. The Clymb, a SailThru customer that sells outdoor gear, saw a 12% increase in email revenue and an 8% increase in total email purchases within 90 days of turning on SailThru’s personalization. After combining personalization with predictions, The Clymb saw a 175% increase in revenue per thousand emails sent, as well as a 72% reduction in churn.

Root cause analysis. Sight Machine, a manufacturing analytics company based in San Francisco and Livonia, Michigan, helps its customers solve complex quality control issues. One problem that Sight Machine’s customers face is interpreting alerts: A quality problem or incident can trigger hundreds of alert codes from potentially thousands of different kinds of sensors along an assembly line. Sight Machine’s software uses machine learning to interpret the patterns of these alerts, helping engineers to quickly pinpoint the few alerts that represent the root cause of the problem, separating them from the ripple effect alerts.

Predictive maintenance. Machine learning can also aid human decision making by discovering meaningful patterns in factory data that people would otherwise be unable to find. Consider Sight Machine again: By analyzing data for patterns that occur before trouble hits, the company’s systems help manufacturing engineers anticipate and prevent problems. For one client deploying a new robotic manufacturing line, Sight Machine was able to reduce downtime by 50% and increase performance by 25% within one month — far better than the 1%–2% performance increases typical of the client’s industry.

It’s still early days for machine-reengineering, so we expect our research to uncover many more new types of machine sidekicks. But it’s already clear that machine-reengineering has the power to help manage the data deluge — and resulting bottlenecks — that modern organizations face. Workers can become more efficient and effective, which improves workflows as well as the bottom line. If data is the path forward, machine-reengineering is paving the way.

Harvard Business Review