The hype surrounding machine learning has been accelerating and expanding for years. Supporters talk about the potential of this technology to improve every process and eliminate any issue. In many cases, the levels of optimism and excitement have reached a fevered pitch.
Less enthusiastic observers, however, have noted that the promise of machine learning is often described in broad terms and abstract assertions. The combination of technical jargon, hazy use cases, and vague details leaves many wondering how accessible and advantageous machine learning really is.
That suspicion is valid, even important. But at the startup level, the machine learning market is innovating and advancing fast. Better still, the technologies in the pipeline are intended neither for the biggest companies nor the most complex processes. Rather, they are being designed to solve age-old business problems that could plague any enterprise.
Machine learning is the cure to these common business headaches.
At some point, most businesses will contend with one of the following issues, and many will grapple with more than one. The good news is that machine learning offers sweeping solutions —set to come online within the next decade — to some of the business world’s most vexing problems:
1. Falling victim to phishing emails.
Phishing attacks are one of the most common and consequential types of cyberscams. In a recent example, phishing emails were used to steal all the W2s from the data storage company Seagate Technology. Once hackers had access to the victims’ SSNs, addresses, and salary info, they filed fraudulent tax returns and stole their refunds.
The single worst phishing scam in history took place when Ubiquiti Networks received a wire transfer request that looked authentic enough that the company processed it without question. Ubiquiti ended up wiring $46.7 million directly into the scammer’s account.
Existing solutions to this problem rely on lagging indicators, meaning they can only spot past schemes but can’t predict future ones. This makes it incredibly easy for scammers to stay one step ahead, and it’s why these schemes continue to be so lucrative.
Agari is one Silicon Valley startup that is using machine learning to flip this security strategy on its head. Instead of studying “bad” emails, it focuses on the much larger number of “good” emails. And since that data set is so much larger, the algorithms the company is developing are that much smarter. This means it will take far less time for the improved technology to raise red flags when phishing emails appear.
Generally, machine learning is the ideal technology for tracking, filtering, and analyzing massive communication ecosystems. In time, the technology will only get better at identifying patterns and spotting anomalies, capabilities that could finally give enterprises an advantage over scammers.
2. Failing to deliver a personalized experience
Consumers expect to receive a highly personalized experience whether they are shopping in person or online. And retailers are eager to deliver it, considering that 71 percent of customers are frustrated when an experience feels impersonal.
Companies like Amazon are great at customizing recommendations and messaging. As a result, 35 percent of what customers purchase on Amazon are the products the company has pushed to them through its catered algorithms. But while this may work for Amazon, a relatively small number of brands are able to deliver this kind of experience, largely due to the technical complexity of machine learning in its current form.
Fortunately, machine learning companies on the cutting edge are working to make the implementation and upkeep of this technology increasingly accessible. Even more exciting, these companies are also working to improve the effectiveness of machine learning’s personalization capabilities.
One such company is CognitiveScale, which offers an out-of-the-box solution with easy implementation thanks to its SaaS platform. It is also one of a few companies pioneering the concept of predictive behavior. By studying patterns of shopping behavior on the individual and demographic level, algorithms are able to anticipate what customers will buy next, even if it’s totally unrelated to their previous purchase.
The applications for this kind of soothsaying are myriad. Looking forward, every e-commerce company, banking platform, and insurance provider will have the means to understand what consumers will do next. In return, consumers will consistently receive the kind of experience that builds engagement and drives conversions.
3. Struggling to train customer service reps
Call centers are notorious for high costs, spotty service, and frequent turnover. Some human representatives’ work is already shifting to chatbots driven by machine learning. Customers, however, still want to talk to a real person.
Machine learning has the potential to address both issues at once. By applying this technology to the training process, it’s possible to institute department-wide standards for excellence. And this applies to both call center reps and to sales teams who lean on phone communication.
The current approach to training relies heavily on real-time monitoring and one-on-one coaching. This method is inefficient and leads to practices and policies that are inconsistently applied. Consequently, performance is degraded by the very process meant to standardize and optimize it.
Among the companies applying machine learning to employee training is Gong, which offers a SaaS platform that plugs into a call center’s phone system and then records, tracks, and analyzes every call. Natural language processing allows trainers to learn what kinds of keywords produce positive and negative outcomes, the ratio of talking to listening, and even what emotions are expressed.
The insights that machine learning produces in this environment reveal what makes good reps successful and how everyone else can emulate their behavior. Once call centers and sales teams begin combining human empathy with data-driven analytics, they will be able to deliver an unparalleled level of service to every future caller.
4. Suffering the effects of an inefficient warehouse
The rise of e-commerce has led the warehouse and distribution industry to grow exponentially over the last two decades. Unfortunately, that growth has not always happened in an optimal way. In 2002, as many as 80 percent of warehouses were operating inefficiently, leading to widespread delays and inaccuracies.
A major retailer like Amazon is able to invest heavily to perfect the fulfillment process, but not every business is able to throw money at the problem. Consequently, delivering the kind of seamless shipping process that customers now expect is a major source of pressure.
Thankfully, companies like 6 River Systems are using machine learning to put fulfillment robots within the reach of smaller enterprises. The company’s robot, “Chuck,” has just four wheels and a touchscreen tablet. Chuck uses mapping software to learn the layout of a fulfillment center, then guides humans to the precise location of products as efficiently as possible. As a result, the size and complexity of sprawling centers no longer lead to confusion and oversights.
As this machine learning application matures in the near future, it will revolutionize fulfillment. Warehouse operators will also experience significant cost savings. More importantly, the kinds of delivery mistakes that alienate customers will be kept to a minimum.
These examples should illustrate just how broadly and deeply machine learning is poised to impact the world of business. Machine learning is not a future technology or something exclusive to mega-corporations. The hype around the technology may be strong, but so far machine learning is meeting and exceeding its promise.
Rama Sekhar is a partner at Norwest Venture Partners focusing on venture investments in enterprise and infrastructure software. Rama’s current investments include Agari, Bitglass, Qubole, SnapRoute, and TRUSTID. He holds an MBA from the Wharton School of the University of Pennsylvania with a double major in finance and entrepreneurial management, as well as a bachelor’s degree in electrical and computer engineering, with high honors, from Rutgers University.