Results promised from artificial intelligence (AI) and machine learning paint a brilliant landscape of possibilities for nimble, data-driven decision making and cost reduction. What do those results look like in reality?
With AI — more specifically, machine learning — companies have been able to create success by presenting and predicting information in a way humans can’t (or can’t afford to) do manually. By detecting and interpreting data across millions, if not billions of data points, machine learning has informed decision makers so they can create solutions that are deeply informed and, in some cases, predictive and proactive.
What does this look like? Recall Netflix, the disruptor of an entire cable television marketplace. Or the NFL, with its razor sharp focus on player insights that no one fathomed just a few years ago.
As businesses commit to invest in and prioritize digital transformation, the potential to make impacts on bottom lines emerges. Companies large and small are leveraging machine learning to automate routine tasks, manage risk, and help people make better decisions. These commitments to data science can not only slash operational budgets, but have the potential to disrupt industries and marketplaces.
Data scientists, the teams that develop machine learning, have propelled companies to next level status by automating operations and surfacing forward-looking intelligence.
Let’s take a look at some big name companies that have leveraged machine learning to make dramatically positive influences in cost reduction, products, services and risk management capabilities.
Machine learning can be used in many ways in the world of advertising. From Lexus’ first ad written by IBM’s Watson to Red Balloon’s discovery of a whole new audience, AI can touch nearly every process of the advertising experience.
RedBalloon, a gift and experience company in Australia, managed to reduce spending per lead acquisition by 30%, with a 3,000% return on some campaigns. How was it possible? The AI model was able to discover micro-audiences that humans at an ad agency simply wouldn’t have time to target (for example, men over 65 in Adelaide who love flying). Advertising dollars went a lot further than the $50 per lead acquired in the past and marketing teams had more free time to make more strategic decisions, while the AI model did the grunt work.
Morningstar, the company that assigns benchmark ratings for investments, has a large team of analysts to examine fund performance. As big as this team is, they weren’t large enough to provide Analyst Ratings for every fund in the U.S.
When Morningstar turned to machine learning to create Quantitative Ratings, the company could review fund performance for six times as many funds. This Quantitative Rating harnesses the power of the wisdom-of-crowds effect. The machine learning algorithm is fed current and historical analyst ratings, and the data used to arrive at those ratings. The final model is a collection of thousands of models that together can create a picture of how an analyst would likely rate a fund.
In the quest to cure cancer, machine learning — combined with biology, statistics, and engineering — saves critical time for medical researchers. Using these combined disciplines, they can sift through billions of DNA sequences and massive amounts of clinical trial data to synthesize and focus on hypotheses that have greater chances of succeeding. Researchers at Bristol Myers Squibb have been using machine learning to generate models that show how therapies interact with the body to predict how medicines could behave in clinical trials. The time savings is huge.
In the past, researchers would study these drug combinations one by one. With next-generation models, they can weed out incompatible pairings, which could save decades of testing. These processes not only generate new data sets, they unlock insights that have never been analyzed before.
When data science meets manufacturing, machine learning models can save money and, in this case, create safer work environments. At Georgia-Pacific, the giant paper manufacturer, data scientists identified a way to reduce paper tears by 40%. They combined data on paper roll quality with the time it takes for paper to tear to create precise schedules for the company’s converting lines. This knowledge allowed Georgia-Pacific to save millions on a single production line, which they can now apply to 150 other lines.
They’ve also learned how to predict equipment failure up to three months in advance. By analyzing machinery performance across time, they have been able to anticipate unplanned downtime, improving asset utilization and safety in their paper mills.
AI and machine learning have transformed the way banks detect — and now predict — fraudulent transactions and money laundering. By combining automated processes with multilayered, deep-learning analysis, companies spend less money to prevent fraud.
Where binary transaction monitoring systems generated warnings with high rates of false positives, AI has reduced those false positives by half. The cost savings also come from focusing human efforts on leads that are less likely to be fraudulent. Where artificial neural networks are used, companies can detect criminals who would otherwise know their way around the binary, rule-based security systems. The result? Banks can predict criminals’ next moves.
Harvard Business Review says AI could become a requirement for fraud detection and prevention for larger businesses because there is “no other way to rapidly detect and interpret patterns across billions of pieces of data.”
For Wells Fargo, it’s not about examining individual transaction points, it’s about observing behaviors along a continuum of transactions. They’re using AI to scour vast amounts of online data, including on the dark web, to identify anti-money laundering signals to make connections humans wouldn’t consider.
The potential for machine learning to help customers aligns with a strong desire for people to seek and discover help on their own, without directly contacting a company. Chat bots and suggested knowledge base algorithms answer this need and can learn from past behaviors to present targeted content at the right time.
If, for example, your customer doesn’t find the answer in a chat bot session, machine learning can track touch points and record their questions so that, if they call you, behind-the-scenes processes quickly unearth appropriate content for the human who answers the phone.
Shortening customer service case referral is always a goal. This is one example where machine learning can help anticipate customer needs, and therefore increase satisfaction. Another is by proactively emailing customers based on their activity on your website. Even better, you can talk to your data scientists about how machine learning can help perform root-cause analysis to pinpoint processes and procedures that need refining and might otherwise go unnoticed.
Overall, when you combine the human touch with automated tasks and machine learning, your company can go a long way toward pleasing your customers and providing — and even anticipating — the services or products they desire.
The next time you watch via a friend’s Netflix account, compare the imagery on the screen to your own. You might notice the same titles displayed, but in a different order, and with different imagery. Stranger Things, for example, could be displayed with an image of teenagers, a spooky landscape with an ominous creature, or a picture of a character’s bloody nose.
Why the difference? Netflix employs user data to personalize the way content is displayed for 140 million products. Machine learning guides each unique view to display titles and images with the goal to anticipate which title you’d like to watch based on your watch history.
As you know, Netflix turned the television industry upside down. They did this, in part, by curating user experiences. Netflix employed machine learning to assign user to data clusters, which considered when programs were watched, users’ ages, genders, times spent watching, and how often they paused and resumed programs.
Along with this data, Netflix assigned more than 76,000 ways to describe movie genres, instead of labelling with traditional (romcom) categories. Why? They wanted to create perfect situations to feed customers content they enjoy. Boom.
It’s clear AI and machine learning can be applied across industries for many types of applications, from process automation to forecasting future performance. And while machine learning can free up time by performing workflow automation, it can also discover deep insights that your human teams might not have time to find. With these additional insights, your company has a new opportunity to anticipate customer needs, understand where new challenges may come from, and have more time to spend on strategic thinking.
Have you thought about how you make machine learning work for your business? We can speak with you about your industry and what it would take for you to start your digital transformation. Contact a member of our team to learn more.