How Your Machine Learning Projects Will Achieve Business Impact

Before long, machine learning will be as omnipresent in our lives as the internet. Like the internet, it will improve every industry and every business. Our grandchildren won’t even know they are using it, and they probably will never appreciate how much it has simplified their lives. The companies that apply this machine learning technology to business problems correctly will transform their organizations by achieving business impact.

Business impact is defined by increasing revenue and/or decreasing cost. So how can you achieve business impact with machine learning? There are three possibilities.

1. Reduce Costs. Machine learning allows you to automate the mundane tasks that drain your time, budget and attention, allowing you to do what you do best. Any activity that requires less than one second of thinking is ripe for being replaced by machine learning. As CEO, I personally led a team that automated more than 21 percent of these roles over two years and the organization has now pushed that number to more than 30 percent in three years. My head of operations did a great job taking all the transaction processing jobs that followed a desktop procedure and replacing them with technology.
2. Gain Efficiency. About 80 percent of the activities that we ask people to do are beneath their skill level. On average only 20 percent make use of their skills. A healthcare example is adjudicating trials. Every participant in a trial needs to be reviewed by a doctor. Let’s say 80 percent of the participants in the trial usually respond as predicted, the doctor still needs to look through all the data and sign off on it. The 20 percent that do not react as planned will require the doctor’s focus and consideration to uncover the cause of the anomaly. By applying machine learning to this space, the technology can quickly identify just the fraction of participants who did not exhibit the predicted response and fast track them for the doctors. These efficiency gains can help a pharmaceutical company save time and money (reduce cost) and bring a drug to market faster (increase revenue).
3. Achieve Breakthroughs. Machine learning unlocks hidden insights and can find “signals” in the data that humans can’t find. Today’s enormous processing power allows machine learning technology to study millions and millions of records almost instantly. Examples of these breakthroughs range from being able to identify guns and knives in luggage, to predicting prostate cancer from a liquid biopsy. Both of these are real examples that can scale to produce major business impact by transforming the thinking of these companies.

But what about the jobs?

There is an unavoidable correlation between business impact and job impact, but the hysteria about artificial intelligence decimating the workforce is exaggerated. When Infinia ML CEO Robbie Allen (then CEO of Automated Insights) helped The Associated Press automatically generate written stories about quarterly corporate earnings, you might have thought the newsroom employees, fearing for their jobs, would revolt.

Instead, The Associated Press reported that “the reaction has been positive from staff, largely because automation has freed up valuable reporting time and reduced the amount of data-processing type work they had been doing.” A New York Magazine headline read that “Robots Are Invading the News Business, and It’s Great for Journalists.” The Verge reported that “computers are not taking journalists’ jobs” and were “freeing up writers to think more critically about the bigger picture.”

---

LinkedIn Co-Founder Reid Hoffman and co-author Chris Yeh share secrets for scaling your business at the CEO2CEO Summit.

Dec. 6, 2018 | New York City | Details and Registration today!

---

Jobs that humans should not perform

That said, the potential for change among workers whose jobs consist of repetitive mindless tasks does exist. Simply put, there are some jobs that humans should not have to do, but until now there has been no other way to get the work done.

Take the example of image analysis for security screening. With sufficient focus people can do this well, but due to the long hours and numerous distractions, performance can degrade. Often it is the very jobs that are critical to important missions like security and health that are most repetitive and vulnerable to human error. By contrast, machine learning doesn’t get distracted and does the same high-quality job at all times. This improved security is in turn a better situation for all people who depend on it.